Harmonicity-dependent controlling of a harmonic filter tool.

The present application refers to the decision on the control of a harmonic filter tool such as the pre / post filter or post-filter approach only. This tool is, for example, applicable to MPEG-D unified audio and voice coding (USAC) and the next 3GPP EVS codec.

Transform-based audio codecs such as AAC, MP3, or TCX generally introduce interharmonic quantization noise when processing harmonic audio signals, particularly at low bit rates.

The effect is even worse when the transform-based audio codec operates with low delay, due to the worse frequency resolution and / or selectivity introduced by a shorter transform size and / or a lower window frequency response.

This inter-harmonic noise in general is perceived as a very annoying twitter element, which significantly reduces the performance of the transformed-based audio codec by subjectively evaluating it in a material of

IX / a / 2017/001240 highly tonal audio such as music or voice.

A common solution to this problem is to use prediction-based techniques, preferably a prediction using a self-regressive (AR) model on the addition or subtraction of input or decoded samples based on,

IX / a / 2017/001240

<td>either</td><td>in the domain of</td><td>transformed or</td><td>in the</td><td>domain of</td>

<td>weather.</td><td></td><td></td><td></td><td></td>

<td>Without</td><td>However, the use of</td><td>these techniques</td><td>in the</td><td>signals with</td>

the changing temporal structure leads again to unwanted effects such as temporary shifting of percussion musical events or strong speeches or even vestiges of impulses due to the repetition of a simple impulse type transient. Therefore, special care must be taken in those signals that contain transient and harmonic components where there is ambiguity between the transients and pulse trains (the latter refers to a harmonic signal composed of very short duration individual pulses; these signals are known as pulses)

There are several solutions to improve the subjective quality of audio codees based on transformed into harmonic audio signals. All of them explore the long-term periodicity (pitch N. of the T. fundamental frequency) of very harmonic, stationary waveforms, and are based on prediction techniques, either in the transformed domain or the time domain. Most solutions are known as long-term prediction (LTP) or pitch prediction, characterized by a pair of filters that are applied to the signal: a pre-filter in the encoder (usually as a first step in the domain of time or frequency) and a post-filter in the decoder (usually as a last step in the time or frequency domain). A few other solutions, however, apply only a simple illiterate post-f process on the decoder side generally known as harmonic post-filter or bass post-filter. All of these approaches, regardless of being pairs of pre- and post-filters or post-filters only, are called the harmonic filter tool.

Examples of transform domains are:

[1] H. Fuchs, Improving MPEG Audio Coding by Backward Adaptive Linear Stereo Prediction, 99th AES Convention, New York, 1995, Preprint 4086.

[2] L. Yin, M. Suonio, M. Váánánen, A New Backward Predictor for MPEG Audio Coding, 103rd AES Convention, New York, 1997, Preprint 4521.

[3] Juha Ojanpera, Mauri Váánánen, Lin Yin, Long Term

IX / a / 2017/001240

Predictor for Transform Domain Perceptual Audio Coding, 107th AES Convention, New York, 1999, Preprint 5036.

Examples of time domain approaches that apply pre- and post-filtering are:

[4] Philip J. Wilson, Harprit Chhatwal, Adaptive transform coder having long term predictor, US Patent 5,012,517, April 30, 1991.

[5] Jeongook Song, Chang-Heon Lee, Hyen-0 Oh, Hong-Goo Kang, Harmonio Enhancement in Low Bitrate Audio Coding Using an Efficient Long-Term Predictor, EURASIP Journal on Advances in Signal Processing, August 2010.

[6] Juin-Hwey Chen, Pitch-based pre-filtering and postfiltering for compression of audio sign, US Patent 8,738,385, May 27, 2014.

[7] Jean-Marc Valin, Koen Vos, Timothy B. Terriberry,

Definition of the Opus Audio Codee, ISSN: 2070-1721,

TETE RFC 6716, September 2012.

[8] Rakesh Taori, Robert J. Sluijter, Eric Kathmann

Transmission System with Speech Encoder with Improved Pitch Detection, US Patent 5,963,895, October 5,

1999.

Examples of time domain approaches that apply only post-filtering are:

IX / a / 2017/001240 [9] Juin — Hwey Chen, Alien Gersho, Adaptive Postfiltering for Quality Enhancement of Coded Speech, IEEE Trans. on Speech and Audio Proc., vol. 3, January 1995.

[10] Int. Telecommunication Union, Erame error robust variable bit — rafe coding of speech and audio from 8-32 kbit / s, Recommendation ITU-T G.718, June 2008.

www.itu.int/rec/T-REC-G.718/e, section 7.4.1.

[11] Int. Telecommunication Union, Coding of speech at 8 kbit / s using conjugate structure algebrare CELP (CSACELP), Recommendation ITU-T G.729, June 2012. www.itu.int/rec/T-REC-G. 729 / e, section 4.2.1.

[12] Bruno Bessette et al., Method and device for frequency-selective pitch enhancement of synthesized speech, US Patent 7,529,660, May 30, 2003.

An example of a transient detector is:

[13] Johannes Hilpert et al., Method and Device for Detecting a Transient in a Discrete-Time Audio Signal, US Patent 6,826,525, November 30, 2004.

Revealing literature in psychoacoustics:

[14] Hugo Fastl, Eberhard Zwicker, Psychoacoustics: Facts and Models, 3rd Edition, Springer, December 14,

2006.

[15] Christoph Markus, Background Noise Estimation,

IX / a / 2017/001240

European Patent EP 2,226,794, March 6, 2009.

All the techniques described decide when to activate the prediction filter based on a single threshold decision (for example prediction gain [5] or pitch gain [4] or harmonicity that is basically proportional to the normalized correlation [6]). In addition, OPUS [7] uses hysteresis that increases the threshold if the pitch is changing and reduces the threshold if the gain in the previous table was above a predefined fixed threshold. OPUS [7] also deactivates the long-term predictor (pitch) if a transient is detected in some specific frame configurations. The reason for this design seems to depart from the general belief that, in a mixture of harmonic and transient signal components, the transient dominates the mix, and the activation of the LTP or pitch prediction would cause, as stated, subjectively more inconveniences that improvements. However, for some waveform mixtures that will be discussed hereinafter, the activation of the long-term predictor or pitch in transient audio frames significantly increases the quality or efficiency of coding and is therefore beneficial. Furthermore, it may be convenient, when activating the predictor, to vary its

IX / a / 2017/001240 intensity based on instantaneous signal characteristics in addition to a prediction gain, the only focus on the prior art.

Accordingly, it is an object of the present invention to provide a concept for the harmonic dependent control of a harmonic filter tool of

X / a / 2017/001240 an audio codec that improves coding efficiency, for example improved objective coding gain or better perceptual quality, or the like.

This object is achieved by claiming in the independent claims of the application.

It is a basic discovery of the present application that the coding efficiency of an audio codec using a controllable - switchable or even adjustable harmonic filter tool can be improved by performing the harmonic dependent control of this tool using a temporal structure measurement in addition to a harmonicity measurement in order to control the harmonic filter tool. In particular, the temporal structure of the audio signal is evaluated so that it depends on the pitch. This makes it possible to achieve a harmonic filter tool control adapted to the situation so that when the control is carried out only based on the harmonicity measurement, the use of this tool is decided against or reduced despite the use of the tool harmonic filter, in such a situation, it would increase the coding efficiency, while in other situations in which the harmonic filter tool may be inefficient or even destructive, The control reduces the application of the harmonic filter tool properly.

The implementations of the present invention on the subject of the dependent claims and their preferred embodiments are described below with reference to the attached Figures.

Figure 1 illustrates a block diagram of an apparatus for controlling a harmonic filter tool in terms of filter gain according to one embodiment;

Figure 2 illustrates an example of a possible predetermined condition to be met to apply the harmonic filter tool;

Figure 3 illustrates a flow chart illustrating a possible implementation of a decision logic that, among other things, could be parameterized to achieve the example condition of Figure 2;

Figure 4 illustrates a block diagram of an apparatus for executing a harmonic-dependent control (and temporal measurement) of a harmonic filter tool;

IX / a / 2017/001240

Figure 5 illustrates a schematic diagram illustrating the temporal position of a temporal region for determining the measurement of temporal structure according to one embodiment;

Figure 6 schematically illustrates a graph of energy samples that temporarily sample the energy of the audio signal within the temporal region according to one embodiment;

Figure 7 illustrates a block diagram illustrating the use of the apparatus of Figure 4 in an audio codec illustrating the encoder and decoder of the audio codec, respectively, when the encoder uses the apparatus of Figure 4, according to an embodiment where a harmonic pre-post filter tool is used;

Figure 8 illustrates a block diagram illustrating the use of the apparatus of Figure 4 in an audio codec illustrating the encoder and decoder of the audio codec, respectively, when the encoder uses the apparatus of Figure 4, according to an embodiment where a harmonic post-filter tool is used;

Figure 9 illustrates a block diagram of the controller of Figure 4 according to one embodiment;

Figure 10 illustrates a block diagram of a system

IX / a / 2017/001240 illustrating the possibility that the apparatus of Figure 4 shares the use of the energy samples of Figure 6 with a transient detector;

Figure 11 illustrates a graph of a time domain portion (waveform portion) of an audio signal as an example of a low pitch signal that also illustrates the positioning according to the pitch of the temporal region to determine the at least unique measurement of temporal structure;

Figure 12 illustrates a graph of a time domain portion of an audio signal as an example of a high pitch signal that also illustrates the positioning according to the pitch of the temporal region to determine the at least single structure measurement temporary;

Figure 13 illustrates an example of a pulse and step transient spectrogram within a harmonic signal;

Figure 14 illustrates an example spectrogram to illustrate the influence of LTP on the impulse and step transient;

Figure 15 illustrates, one above the other, the time domain portions of the audio signal illustrated in Figure 14, and their version with low and high pass filter, respectively, in order to illustrate the control of agreement

IX / a / 2017/001240 to Figure 2, 3, 16 and 17 for the impulse and step transient;

Figure 16 illustrates a bar graph of an example for the temporal sequence of energy of segments of the energy sample sequence - for a pulse transient and the arrangement of the temporal region to determine the at least single measurement of temporal structure according to Figure 2 and 3;

Figure 17 illustrates a bar graph of an example for the temporal sequence of segment energy sequence energy samples - for a transitory passage and the arrangement of the temporal region to determine the at least single measurement of temporal structure according to Figure 2 and 3;

Figure 18 illustrates an example of a pulse train spectrogram (extract using a short FFT spectrogram);

Figure 19 illustrates an example of a pulse train waveform;

Figure 20 illustrates an original short FFT spectrogram of the pulse train; Y

Figure 21 illustrates an original long FFT spectrogram of the pulse train.

The following description begins with a first detailed description of the control of a harmonic filter tool. The ideas that led to this first realization are briefly described. These ideas, however, also apply to other embodiments. Then, general embodiments are presented followed by specific concrete examples of portions of audio signals with the

IX / a / 2017/001240

<td>end to delineate of</td><td>a way</td><td>plus</td><td>concrete</td><td>the</td><td>effects</td><td>what</td>

<td colspan="2">result from the accomplishments</td><td>of the</td><td>request</td><td> •</td><td></td><td></td>

<td>The mechanism of</td><td>decision</td><td>for</td><td>Activate</td><td>or</td><td>control</td><td>a</td>

Harmonic filter tool of, for example, a prediction-based technique, is based on a combination of a harmonicity measurement such as a normalized correlation or prediction gain and a temporal structure measurement, for example a temporal flatness measurement or energy change

The decision may, according to the foregoing, not only depend on the harmonic measurement of the current table, but also on the previous table, and on a temporal structure measurement of the current table and optionally the previous table.

The decision scheme can be designed so that the prediction-based technique is also activated for transients, whenever its use is psychoacoustically beneficial according to a respective model.

The thresholds used to activate the prediction-based technique may, in one embodiment, be dependent on the current pitch instead of the pitch change.

The decision scheme allows, for example, to avoid the repetition of a specific transient, but allows the technique based on the prediction for some transients and for signals with specific temporal structures where a transient detector would normally signal short transform blocks (i.e., the existence of one or more transients).

The decision technique presented below can

IX / a / 2017/001240

<td>be applied to</td><td>anyone</td><td colspan="2">of the methods</td><td>based</td><td>in</td>

<td>prediction before</td><td>described,</td><td>either in</td><td>the</td><td>domain</td><td>from</td>

<td>transformed or in</td><td>The Dominion</td><td>specific already</td><td>be</td><td colspan="2">in approaches</td>

Pre-filter plus post-filter or post-filter only. On the other hand, it can be applied to predictors that operate limited by bands (with low pass) or in sub-bands (with band pass characteristics).

The general objective related to the activation of LTP, pitch prediction, or harmonic post filtration is the achievement of both conditions, namely:

obtain an objective or subjective benefit by activating the filter, the non-introduction of significant artifacts by activating said filter.

The determination of the benefit of using the filter is usually performed by means of auto correlation measurements and / or prediction gain in the given signal and is well known [1-7].

The measurement of a subjective benefit is also direct at least for the stationary signals, since the perceptual improvement data obtained by the listening tests are typically proportional to the corresponding objective measurements, that is, the correlation gain and / or prediction before mentioned.

The identification or prediction of the existence of artifacts through filtering, however, requires more sophisticated techniques than simple comparisons of objective measurements such as type of frames (long transforms for stationary frames vs. short transforms for transient frames) or prediction gain for certain thresholds, as is done in the art. Essentially, in order to prevent artifacts, it must be ensured that the changes caused by filtering in the target waveform do not significantly exceed a spectrum threshold

IX / a / 2017/001240 time variable time at any point in time or frequency. The decision scheme according to some of the embodiments presented below, consequently, uses the following decision and control scheme comprising three algorithmic blocks to be executed in series for each frame of the audio signal to be encoded and / or submitted to the filtered out:

A harmonicity measurement block that calculates harmonic filter data commonly used as normalized correlation or gain values ​​(hereinafter prediction gain). As indicated again, the word gain is a generalization for any parameter commonly associated with the intensity of the filter, for example, an explicit gain factor or the absolute or relative magnitude of a set of one or more filter coefficients.

A T / F envelope measurement block that computes time-frequency amplitude (T / F) or energy or flatness data with a predefined spectral and temporal resolution (this may also include transient measurements of the frame used for

IX / a / 2017/001240 the decisions by cadres, as indicated). The pitch obtained in the harmonicity measurement block is entered into the T / F envelope measurement block since the region of the audio signal used to filter the current frame, typically using past signal samples, depends on the pitch ( the same happens with the computed T / F envelope).

A block of computation of the filter gain makes the final decision on which filter gain to use for filtering (and transmitting in the bit rate). Ideally, this block should compute, for each transmissible filter gain less than or equal to the prediction gain, a standard-type spectrum-time excitation envelope of the target signal after filtering with said filter gain, and comparing this current envelope with an excitation pattern envelope of the original signal. It is then possible to use the higher filter gain for the encoding / transmission whose current envelope corresponding time spectrum does not differ from the original envelope by more than a certain measurement. This filter gain will be called psychoacoustically.

IX / a / 2017/001240 optimal.

IX / a / 2017/001240

In other embodiments described below, the three block structure is somewhat modified.

In other words, the harmonicity and T / F envelope measurements are obtained in the corresponding blocks that are subsequently used to derive the psychoacoustic excitation patterns of both input and output, and finally the filter gain is adapted so that a threshold of masking, given by a relationship between the current and the original envelope does not significantly exceed. To appreciate this, it should be noted that an excitation pattern in this context is very similar to a spectrogram-like representation of the examined signal, but that it exhibits a temporal attenuation modeled from certain characteristics of human hearing and manifests as post- masking

Figure 1 illustrates the connection between the three blocks presented above. Unfortunately, a derivation in the sense of two excitation patterns and a brute force search for the best filter gain is often computationally complex. Consequently, simplifications are presented in the following description.

In order to avoid expensive computations of the excitation patterns in the proposed filter activation decision scheme, low complexity envelope measurements are used as estimates of the characteristics of the excitation patterns. It has been found that in the T / F envelope measurement block, data such as segmental energies (SE), temporal flatness measurement (TFM), maximum energy change (MEC) or traditional frame configuration information such as type The table (long / fixed or short / transient) is sufficient to derive estimates of psychoacoustic criteria. These estimates can then be used in the computation block of the filter gain to determine, with high precision, an optimum filter gain to be used for coding or transmission. In order to avoid a computationally intense search for the globally optimal gain, a rate distortion loop over all possible gains (or a subset thereof) can be replaced by unique conditional operators. These cheap operators are used to decide whether the filter gain, computed using data from the T / F harmonicity and envelope measurement blocks, will be zero (decision not to use harmonic filtering) or not (decision to use filtering

IX / a / 2017/001240 harmonic). Note that the harmonicity measurement block can remain unchanged. A step-by-step implementation of this low complexity embodiment is described below.

As indicated, the initial filter gain submitted to the unique conditional operators is derived using data from the T / F harmonicity and envelope measurement blogues. More specifically, the initial filter gain can be equal to the product of the prediction gain variable over time (from the harmonicity measurement block) and a variable scale factor by time (from the psychoacoustic envelope data of the T / F envelope measurement block). In order to further reduce the computational load, a constant, fixed scale factor such as 0.625 can be used instead of the variable with the time adaptable to the signal. This typically retains sufficient quality and is considered in the following embodiment.

A step-by-step description of a specific embodiment to control the filter tool is set forth below.

IX / a / 2017/001240

one. Transient detection and temporary measurements et no

The input signal s_HP(n) enters the non-transient detector by time domain. The input signal s_HP(n) a step is submitted

The transfer function of the high-pass detection is

H_td(z) = 0.375-0.5Γ¹ + 0.125z "²

The signal filtered by the high-pass detection filter is indicated as s_rD(n) · The Signal undergoing a segmented pass filter into eight consecutive segments of the same length.

The energy of the signal subjected to high-pass filtering each segment is calculated as follows:

L $ egmen {1 'segment i = 0, ..., 7 n-0 where L_segment= ^ is the number of samples in a segment of

2.5 milliseconds at the input sampling frequency.

The accumulated energy is calculated using:

^andAcc = max (E_rD(z -1), 0.8125E_Acc)

An attack is detected if the energy of a segment E_TD(i) exceeds the energy accumulated by a constant factor attackRatb = 8.5 and the attacklndex is set at i:

Ejp (zj&gt; attackRatio E_Acc ( 4 )

If no attack is detected based on the preceding criteria, but a strong increase in energy is detected in segment i, the attack index is fixed on it without indicating the presence of an attack. The attacklndex basically set itself to the position of the last attack in a box with some additional restrictions. The energy change for each segment is calculated as follows:

^ANDchng ^ANDTD (0

<img file="MX366278B_D0001.tif" />

£_rD(zj

IX / a / 2017/001240 ^ANDTD ^)&gt; ^ANDTD ^ ^andtd (í ^_1) &gt; ^£tdG) (5)

The measurement of temporal flatness is calculated as follows:

(6) ¡-Μ ^{one 1 v} past

The maximum energy change is calculated as follows:

MEC [N_past,

Nnew) past \ ^ chng {. past ^^ '“f ^ chng ^ new 1)) (7)

If the index of E_chng(i) or E_TD(i) is negative then indicates a value of the previous segment, advancing the segment in relation to the current table.

N_pastIt is the number of the segments of the last frames. It is equal to 0 if the temporal flatness measurement is calculated for use in the ACELP / TCX decision. If the temporal flatness measurement is calculated for the TCX LTP decision then it is equal to:

^2V past = l + min 8, 8 ^ + 0.5 l L (8)

IX / a / 2017/001240

N_n,,_wIt is the number of segments of the current frame. It is equal to 8 for non-transitory tables. For the transitional tables, the locations of the segments with the maximum and minimum energy are first established:

* max = argmax E_TD (J ^and{- / V_{paJ (} , ..., 7) (9) et no ¡min = argmin E_TD(i) ^andl-fV_paJ, ,...,7}

If E_td (¡Mili)&gt; O.375E_rD (i_raax) then N_mis fixed at no max

-3, otherwise set to 8.

two. Change of length of the transform block

The overlap length and the length of the TCX transform block depend on the existence of a transient and its location.

Table 1: Overlay coding and transform length based on transient position

<td>IIIIIIIIIIII</td><td></td><td>i | y ||||||| IÍiÉiOO ^ ®úd ||| yÍí | ó | íll BooPPIOOOOOOOOÑÓÓóOs ^ hi ^ hhhhhhhhhyOOOOOOOOÑOO</td><td>llllllielyiilllll |||| and || eáOÍ ||||| IIIIO | l |}: § | y | ^ lll | lSSgyéí ^ pSÍeil: llllllllpiillllllll</td><td>||||||| ί || 1 |} ύ ||| from ll ^ l ^ ogoil ition</td>

<td>No or</td><td>ALDO</td><td> 0</td><td> 0</td><td> 00</td>

<td> -2</td><td>TOTAL</td><td> 1</td><td> 0</td><td> 10</td>

<td> -1</td><td>TOTAL</td><td> 1</td><td> 0</td><td> 10</td>

<td> 0</td><td>TOTAL</td><td> 1</td><td> 0</td><td> 10</td>

<td> 1</td><td>TOTAL</td><td> 1</td><td> 0</td><td> 10</td>

<td> 2</td><td>MINIMUM</td><td> 1</td><td> 10</td><td> 110</td>

<td> 3</td><td>INTERMEDIATE</td><td> 1</td><td> 11</td><td> 111</td>

<td> 4</td><td>INTERMEDIATE</td><td> 1</td><td> 11</td><td> 111</td>

<td> 5</td><td>MINIMUM</td><td> 1</td><td> 10</td><td> 110</td>

<td> 6</td><td>MINIMUM</td><td> 0</td><td> 10</td><td> 010</td>

<td> 7</td><td>INTERMEDIATE</td><td> 0</td><td> 11</td><td> 011</td>

IX / a / 2017/001240

The transient detector described above basically returns the index of the last attack with the restriction that if there are multiple transients then a MINIMUM overlay is preferred over an INTERMEDIATE that is preferred over a TOTAL overlay. If the attack in position 2 or 6 is not strong enough then the INTERMEDIATE overlay is chosen instead of the MINIMUM.

3. Pitch estimation

A pitch delay (whole part + fraction part) per frame (frame size for example 20ms) is estimated. This is done in three stages to reduce complexity and improve the accuracy of the estimate.

to. First estimate on the whole part of the pitch delay

A pitch analysis algorithm is used that produces a smooth pitch evolution profile (for example the Open-loop pitch analysis described in ITU-T Rec. G.718, section 6.6). This analysis in general is carried out by sub-tables (subframe size for example lOms), and produces an estimate of

IX / a / 2017/001240

<td>delay of</td><td>pitch</td><td>by</td><td>subframe</td><td>Notice</td><td>what</td><td>these</td>

<td>estimate</td><td colspan="2">delay</td><td>pitch</td><td>they do not have</td><td>a</td><td>part</td>

<td>fractional</td><td>and in</td><td colspan="3">general are estimated over</td><td>a</td><td>signal</td>

<td>sampled</td><td>toward</td><td>down</td><td>(Cup of</td><td>sampling,</td><td>by</td><td>example</td>

6400Hz) The signal used can be any audio signal, for example an LPC compensated audio signal as described in ITU-T Rec. G.718, sec. 6.5.

b. Refinement of the entire part of the pitch delay

The final integer part of the pitch delay is estimated on an audio signal x [n] that runs at the sampling rate of the central encoder, which is generally higher

<td>to that of the</td><td>sampled signal</td><td>down</td><td>used</td><td>in a.</td><td>(by</td>

<td>example 12.8kHz</td><td>, 16kHz, 32kHz ...</td><td>). The signal</td><td>x [n]</td><td>may</td><td>be</td>

<td>any signal</td><td>audio by</td><td>example one</td><td>signal</td><td colspan="2">audio</td>

LPC offset.

The whole part of the pitch delay is then the delay

Tmt that maximizes the auto-correlation function

L

C (d) = x [n] x [n - d] n = 0

IX / a / 2017/001240 with d around an estimated pitch delay T in the step

1st

c. Estimation of the fractional part of the pitch delay

The fractional part is interpolating the auto correlation function C (d) computed in step 2.b. and selecting the fractional pitch delay Tf_r which maximizes the interpolated auto correlation function. Interpolation can be performed with a low pass filter FIR as described for example in Rec. ITU-T G.718, sec. 6.6.7.

Four. Decision bit

If the input audio signal does not contain harmonic content or if a prediction-based technique would introduce distortions in the temporal structure (for example repetition of a short transient), then no parameters are encoded in the bit stream. Only 1 bit is sent so that the decoder knows whether to decode the filtering parameters or not. The decision is made based on several parameters.

The normalized correlation in the integer pitch delay estimated in step 3.b.

Ση = 0 * Μ * [^ - ^ ηί] norm_corr = νΣη = ο * ΐΑΐ^χΐΛΐ νΣη = ο * [^η - ^r¿Nt] ^ [ⁿ “ ^Tint]

The normalized correlation is 1 if the input signal is perfectly predictable for an integer pitch delay, and 0 if it is not predictable. A high value (close to 1) would indicate a harmonic signal. For a more precise decision, in addition to the normalized correlation for the current table (norm_corr (curr)) the normalized correlation of the last table (norm_corr (prev)) can also be used in the decision, for example:

Yes (norm_corr (curr) * norm_corr (prev))&gt; 0.25 or

If max (norm_corr (curr), norm_corr (prev))&gt; 0.5, then the current frame contains some harmonic content (bit = l)

to. Features computed by a transient detector (for example Temporary flatness measurement (6), Maximum energy change (7)), to prevent activation of the post filter in a signal that contains a strong transient or large temporal changes. The temporal characteristics are calculated on the signal that contains the current frame (N_new segments) and the frame passed until the pitch delay (N_past segments). For transients that are slowly decaying, all or some of the characteristics are calculated only to the position of the transient (i_m„-3) since distortions in the non-harmonic part of the spectrum introduced by the LTO filtering would be suppressed by prolonged transient masking (for example a cymbal crash cymbal).

b. Pulse trains for low pitch signals can be detected as a transient by a transient detector. For signals with low pitch the characteristics of the transient detector are ignored and there is instead an additional threshold for the normalized correlation that depends on the pitch delay, for example:

If norm_corr &lt;= 1.2-T¿_nt/ L, then bit = 0 is set and

IX / a / 2017/001240 parameters are not sent.

An example decision is illustrated in Figure 2 where bl is some bit rate, for example 48 kbps, where TCX_20 indicates that the frame is encoded using a simple long block, where TCX_10 indicates that the frame is encoded using 2.3, 4 or more short blocks, where decision TCX_20 / TCX_10 is based on the output of the transient detector described. tempFlatness is the Time Planarity Measurement defined in (6), maxEnergyChange is the Maximum Energy Change defined in (7). The norm_corr (curr)&gt; 1.2-T¿ condition_nt/ L could also be written as (1.2-norm_corr (curr)) * L &lt;T_int.

The principle of decision logic is illustrated in the block diagram in Figure 3. It should be noted that Figure 3 is more general than Figure 2 in the sense that the thresholds are not limited. They can be fixed according to Figure 2 or in a different way. On the other hand, Figure 3 illustrates that the example of bit rate dependence of Figure 2 can be set aside. Naturally, the decision logic of Figure 3 could be modified to include the bit rate dependence of Figure 2. In addition, Figure 3 has been maintained nonspecific with respect to the use of only the current or also the past pitch. In

IX / a / 2017/001240 therefore, Figure 3 illustrates that the embodiment of Figure 2 can be modified in this regard.

The threshold in Figure 3 corresponds to different thresholds used for tempFlatness and maxEnergyChange in Figure 2. The threshold_l in Figure 3 corresponds to 1.2T¿_nt/ L in Figure 2. The threshold_2 in Figure 3 corresponds to 0.44 or max (norm_corr (curr), norm_corr (prev))&gt;

0.5 or (norm_corr (curr) * norm_corr_prev)&gt; 0.25 in Figure 2.

It is obvious from the preceding examples that the detection of a transient affects the decision mechanism for long-term prediction and what part of the signal will be used for the measurements used in the decision, and does not directly trigger the deactivation of the prediction long-term.

The temporary measurements used for the transform length decision may be completely different from those used for the LTP decision or overlap or be exactly the same although calculated in different regions.

For low pitch signals the transient detection is completely ignored if the threshold for the normalized correlation that depends on the pitch delay is

IX / a / 2017/001240 reached.

5. Estimation and quantization of profit

The overall gain is estimated on the input audio signal at the sampling rate of the central encoder, but it can be any audio signal such as the LPC Compensated Audio Signal. This signal is indicated as y [n] and can be the same or different from x [n].

The prediction y? [N] of y [n] is first filtered and [n] with the following filter

IX / a / 2017/001240

P (z ') = B (z, Tf_r) z ^Tint where T_nt is the integer part of the pitch delay (estimated at 0) and a low pass FIR filter whose coefficient depends on the fractional part of the pitch delay Tf_r(estimated at 0).

An example of B (z) when the pitch delay resolution is

<img file="MX366278B_D0002.tif" />

<img file="MX366278B_D0003.tif" />

B (z) = O.OOOOz-² + 0.2325z^_1 + 0.5349z ° + 0.2325Z¹

B (z) = 0.0152z^-2 + 0.3400z^_1 + 0.5094z ° + 0.1353Z¹

T_fr = - B (z) = 0.0609z^-2 + 0.4391z^_1 + 0.4391z ° + 0.0609Z¹

T_fr = - B (z) = 0.1353z^-2 + 0.5094z^_1 + 0.3400z ° + 0.0152Z¹

IX / a / 2017/001240

The gain g is then computed as follows:

_ Zn = oy [n] and_P[n] ⁹ ZíüoypWypfa] and is limited between 0 and 1.

Finally, the gain is quantified, for example, in 2 bits, using for example uniform quantization.

If the gain is quantized at 0, then no parameter is encoded in the bit rate, only the decision bit 1 (bit = 0).

The foregoing description whereby the advantages of the embodiments herein have been indicated for a harmonic-dependent control of a harmonic filter tool, also applies to the general embodiments that follow. In some cases the foregoing description has been very specific although the concept of harmonicity-dependent control can also be applied within the framework of other audio codes and can be modified with respect to the details set forth above. For this reason, the embodiments of the present application will be described again in a more generic manner. However, the following description may refer to the preceding detailed description with explaining the implementation of the general embodiments. To this end, it is possible to transfer certain implementation details of the above description to the elements described below. Consequently, whenever the description that follows refers to the preceding one, it will be understood that it is independent of the additional references thereto.

Thus, a more generic embodiment that emerges from the previous detailed description is illustrated in Figure 4. In particular, Figure 4 illustrates an apparatus for executing a harmonic dependent control of a harmonic filter tool, such as a filter tool. Harmonic pre / post or post-filter, of an audio codec. The device in general is indicated by reference 10. The apparatus 10 receives the audio signal 12 to be processed by the audio codec and issues a control signal 14 to fulfill the control task of the apparatus 10. The apparatus 10 comprises a pitch estimator 16 configured to determine a pitch delay current 18 of the audio signal 12, and a harmonicity meter 20 configured to determine a measurement 22 of the harmonicity of the audio signal 12 using a current pitch delay 18. In particular, the harmonicity measurement can be a prediction gain or can be implemented by one or more filtering coefficients (single or multi-tap) or a maximum normalized correlation. The calculation block of the harmonicity measurement of Figure 1 comprises the tasks of the pitch estimator 16 and harmonicity meter 20.

The apparatus 10 further comprises a temporal structure analyzer 24 configured to determine at least one temporal structure measurement 26 according to pitch delay 18, the measurement 2 6 that measures a characteristic of a temporal structure of the audio signal 12. For example, the dependence may be based on the positioning of the temporal region within which the measurement 26 measures the characteristic of a temporal structure of the audio signal 12, as described and will be described in greater detail below. In honor of completeness, however, it should be noted that the dependence of the measurement determination 26 on the pitch delay 18 can be carried out in a different way from that described above and subsequently. For example, instead of positioning the temporary portion, that is, the window

IX / a / 2017/001240 of determination, according to the pitch delay, the dependence could merely temporarily vary the weights at which a respective time interval of the audio signal within a window arranged independently of the pitch delay in relation to to the current picture, they contribute to measurement 26. In relation to the description that follows, this may mean that the determination window 36 could be arranged steadily to correspond to the concatenation of the current and previous frames, and that the portion arranged relative to the pitch merely functions as a window of a greater weight to which the temporal structure of the audio signal influences measurement 26. However, for the moment, it is assumed that the time window is arranged according to the pitch delay. The temporal structure analyzer 24 corresponds to the calculation block of the T / F envelope measurement of Figure 1.

Finally, the apparatus of Figure 4 comprises a controller 28 configured to emit a control signal 14 according to the temporal structure measurement 2 6 and the harmonicity measurement 22 to consequently control the harmonic pre / post filter or the harmonic post filter. When you compare Figure 4 with Figure 1, the

IX / a / 2017/001240 computing block of the optimal filter gain corresponds to or represents a possible implementation of controller 28.

The mode of operation of the apparatus 10 is as follows. In particular, the function of the apparatus 10 is to control the harmonic filter tool of an audio codec, and despite the fact that the foregoing more detailed description with respect to Figures 1 to 3 reveals a gradual control or adaptation of this tool in terms of its filter intensity or filter gain, for example, controller 28 is not limited to that kind of gradual control. In general, the control by the controller 28 can gradually adapt the filter intensity or gain of the harmonicity filter tool between 0 and a maximum maximum value, including, as in the case of the specific examples above with respect to the Figures 1 to 3, although different alternatives are possible, such as a gradual control between two non-zero filter gain values, a progressive control or a binary control such as switching between activation (not zero) and deactivation (zero gain) to turn the harmonic filter tool on and off.

As is clear from the above exposure, the harmonic filter tool illustrated in the

IX / a / 2017/001240

Figure 4 by dotted lines 30 is intended to improve the subjective quality of an audio codec as a transformed-based audio codec, especially with respect to the harmonic phases of the audio signal. In particular, this tool 30 is especially useful in low bit rate scenarios where the quantization noise introduced, without the tool 30, would lead such harmonic phases to audible artifacts. It is important, however, that the filter tool 30 does not adversely affect other time phases of the audio signal that are not predominantly harmonic. In addition, as indicated, the filter tool 30 may be of the post-filter or pre-filter approach. Pre and post-filters can operate in the transformed domain or time domain. For example, a post-filter of the tool 30 can, for example, fulfill a transfer function with a maximum local that is available at spectral distances that correspond to, or are set according to the pitch delay 18. The implementation of a Pre-filter and / or post-filter in the form of an LTP filter, in the form of, for example, a FIR and IIR filter, respectively, is also possible. The pre-filter may have a transfer function that is substantially inverse to the transfer function of the

IX / a / 2017/001240 post-filter. In effect, the pre-filter seeks to hide the quantization noise within the harmonic component of the audio signal by increasing the quantization noise within the harmonic of the current pitch of the audio signal and the postfilter reformulates the transmitted spectrum accordingly. In the case of the post-filter approach only, the post-filter

IX / a / 2017/001240

<td colspan="2">modify</td><td>the signal of</td><td>Audio</td><td>transmitted to filter</td><td>noise</td>

<td>from</td><td>quanti</td><td>.zation that</td><td>it happens</td><td>between the harmonics of</td><td>pitch of</td>

<td>the</td><td>signal</td><td>audio</td><td></td><td></td><td></td>

<td></td><td>Should</td><td>warn</td><td>that</td><td>Figure 4 is, in some</td><td>sense,</td>

conceived in a simple way. For example, although Figure 4 suggests that the pitch estimator 16, the harmonicity meter 20 and the temporal structure analyzer 24 operate, that is, perform their functions, on the audio signal 12 directly, it does not it is necessary. Instead, the pitch estimator 16, the temporal structure analyzer 24 and the harmonicity meter 20 can operate on different versions of the audio signal 12 such as those other than the original and some pre-modified version thereof, as long as These versions may vary between elements 16, 20 and 24 internally and also in relation to the audio codec as well, and may also operate on some modified version of the original audio signal. For example, the temporary structure analyzer 24 can operate

<td>on</td><td colspan="2">signal</td><td>audio 12 to</td><td colspan="2">speed of</td><td colspan="2">sampling of</td>

<td>entry</td><td>from</td><td>the</td><td>same, that is</td><td>the</td><td>speed</td><td>from</td><td>sampling</td>

<td>original</td><td>from</td><td>the</td><td>audio signal 12,</td><td>or</td><td colspan="2">can operate</td><td>over one</td>

IX / a / 2017/001240 internally encoded / decoded version thereof. The audio codec, in turn, can operate at an internal central sampling rate that is usually lower than the internal sampling rate. The pitch estimator 16, in turn, can perform its pitch estimate on a pre-modified version of the audio signal, such as, for example, a psychoacoustically compensated version of the audio signal 12 in order to improve the estimate. pitch with respect to spectral components that are, in terms of perceptibility, more significant than other spectral components. For example, as described, the pitch estimator 16 may be configured to determine the pitch delay 18 in stages comprising a first stage and a second stage, the first stage results in a preliminary estimate of the pitch delay which then It is refined in the second stage. For example, as described, the pitch estimator 16 can determine a preliminary estimate of the pitch delay in a domain sampled down that corresponds to a first sampling rate, and then refining the preliminary estimate of the pitch delay to a second sampling rate that is greater than the first sampling rate.

With regard to the harmonicity meter 20, it is clear

IX / a / 2017/001240

<td>from</td><td>The exhibition</td><td>preceding</td><td>with regard</td><td>to</td><td>the</td>

<td>Figures 1 to</td><td colspan="2">3 that can determine the</td><td>measurement 22</td><td>from</td><td>the</td>

<td>harmonicity</td><td>computing a</td><td>correlation</td><td>normalized</td><td>from</td><td>the</td>

audio signal or a pre-modified version of it in the

<td>delay of</td><td colspan="3">pitch 18. It should be noted that</td><td>the</td><td>measurer</td><td>from</td>

<td>harmonicity</td><td colspan="4">20 can still be set to</td><td>compute</td><td>the</td>

<td>correlation</td><td>normalized yet</td><td>to</td><td>several</td><td colspan="2">distances</td><td>from</td>

<td>correlation</td><td>in addition to the delay</td><td>from</td><td>pitch</td><td> 18</td><td>like in</td><td>a</td>

<td>interval c</td><td>the temporary delay</td><td>what</td><td>It includes</td><td>Y</td><td>go around</td><td>the</td>

<td>delay of</td><td>pitch 18. This may</td><td>be</td><td colspan="2">favorable,</td><td colspan="2">for example,</td>

in the case of a filter tool 30 that uses a multi-tap LTP or possible LTP with fractional pitch. In that case, the harmonicity meter 20 can analyze or evaluate the correlation even at delay rates close to the current pitch delay 18, such as the entire pitch delay in the preceding concrete example set forth in Figures 1 to 3.

For details and possible additional implementations of the pitch 16 estimator, refer to the preceding pitch estimation section. The possible implementations of the harmonicity meter 20 were discussed above with respect to the norm.corr equation. However, also as described, the expression harmonicity measurement will not only include a normalized correlation but also suggests the prediction gain of the harmonic filter, while the harmonic filter may or may not be equal to the pre-filter of the 230 filter in case of using in pre / post-filter approach regardless of the audio codec that this harmonic filter uses or if this harmonic filter is merely used by the harmonic meter 20 to set the measurement 22.

As in the case of Figures 1 to 3 described above, the temporal structure analyzer 24 may be configured to determine the at least single measurement of temporal structure 26 within a temporal region temporarily placed according to pitch delay 18. To For greater illustration, see Figure 5. Figure 5 illustrates a spectrogram 32 of the audio signal, that is, its spectral decomposition to its highest frequency f_H according to, for example, the sampling rate of the version of the audio signal internally used by the temporal structure analyzer 24, temporarily sampled at a transform block rate that

IX / a / 2017/001240 may or may not match the transform block rate of the audio codec, if applicable. For illustrative purposes, Figure 5 shows spectrogram 32 as temporarily subdivided into frames in units with which the controller can control the filter tool 30, the subdivision can, for example, also coincide with the subdivision of frames used by the audio codec that comprises or uses the filter tool 30.

At the moment, for the purposes of the illustration it is assumed that the current frame for which controller 28 executes the control, is frame 34a. As described and as illustrated in Figure 5, the temporal region 36, within which the structure analyzer determiner

IX / a / 2017/001240

<td>temporary</td><td>determines</td><td>the</td><td>by</td><td colspan="2">the least unique</td><td>measurement</td><td>from</td>

<td colspan="2">temporary structure</td><td> 26,</td><td>do not</td><td>necessarily</td><td colspan="2">coincides with</td><td>the</td>

<td>picture</td><td>current</td><td>34a.</td><td>In</td><td>change both</td><td>the</td><td>extreme</td><td>from</td>

Temporarily past header 38 as the temporarily future header end 40 of the temporal region 36 may depart from the temporarily past and future header ends 42 and 44 of the current frame 34a. As described, the temporal structure analyzer 24 may arrange the temporarily past header end 38 of the temporal region 36 according to the pitch delay 18 determined by the pitch estimator 16 which sets the pitch delay 18 for each frame 34 , by current table 34a. As will be noted from the foregoing description, the temporary structure analyzer 24 may arrange the temporarily past header end 38 of the temporary region such that the temporarily passed header end 38 moves toward the past relative to the header end temporarily past 42 of the current frame 34a, for example, in a degree 4 6 that monotonic increases with the increase in pitch delay 18. In other words, the greater the pitch delay 18, the greater will be 46. As will be noted from the preceding discussion with respect to Figures 1 to 3, the degree can be set according to equation 8, where N_past it is a measurement for temporal displacement 46.

The temporarily future header end 40 of time region 36, in turn, can be fixed by the time structure analyzer 24 according to the time structure of the audio signal within a time candidate region 48 extending from the end from temporarily passed header 38 of temporal region 36 to the temporarily future header end of the frame

IX / a / 2017/001240, 44. In particular, as stated above, the temporal structure analyzer 24 can evaluate a measurement of energy sample disparity of the audio signal within the temporal candidate region 48 to decide on the position of the temporarily future header end 40 of the temporal region 36. In the specific previous details presented with respect to Figures 1 to 3, a measurement of the difference between the samples of maximum and minimum energy within the temporal candidate region 48 was used as a measure of disparity, as a ratio of amplitude between them. In particular, in the previous concrete example, the variable N_new measures the temporarily future header end 40 of the temporarily future 36 with respect to the temporarily passed header end 42 of the current frame 34a as indicated in 50 in Figure 5.

As will be noted from the above exposure, the arrangement of the time region 36 dependent on the pitch delay 18 is convenient for the ability of the apparatus 10 to correctly identify situations where the harmonic filter tool 30 can be conveniently used. In particular, the correct detection of such situations is more reliable, that is, such situations are more likely to be detected without substantially increasing the detection of false positives.

As described with respect to Figures 1 to 3, the temporal structure analyzer 24 can determine the at least unique measurement of temporal structure within the temporal region 36 on the basis of a temporary sampling of the energy of the signal of audio within that time region 36. This is illustrated in Figure 6, where energy samples are indicated by points plotted on a time / energy plane encompassed by arbitrary time and energy axes. As explained, the energy samples 52 may have been obtained by mastering the energy of the audio signal at a sampling rate greater than the frame rate of the frames 34. In determining the at least one measurement of temporal structure 26, the analyzer 24 may, as stated, compute for example a set of energy change values ​​during a change between the pairs of immediately consecutive energy samples 52 within of the temporal region 36. In the preceding description, equation 5 was used for this purpose. By this measurement, an energy change value can be obtained from each pair of immediately consecutive energy samples 52. The analyzer 24 can

IX / a / 2017/001240 then subject the set of energy change values ​​obtained from the energy samples 52 within the temporal region 36 to a scalar function to obtain the at least single measurement of structural energy 26. In the example In particular, the measurement of temporal flatness, for example, has been determined on the basis of a sum of addends, which depend respectively on exactly a set of energy change values. The maximum energy change, in turn, was determined according to equation 7 using a maximum operator applied over the energy change values.

As indicated, the energy samples 52 do not necessarily measure the energy of the audio signal 12 in its original unmodified version. Instead, energy sampling 52 can measure the energy of the audio signal in some modified domain. In the preceding concrete example, for example, the energy samples measure the energy of the audio signal obtained after high pass filtering thereof. Consequently, the energy of the audio signal in a spectrally lower region influences the samples of

IX / a / 2017/001240

<td>Energy</td><td> 52</td><td>less</td><td>what</td><td>the</td><td>components</td><td>spectrally</td>

<td>superior</td><td>from</td><td>the</td><td>signal</td><td>from</td><td colspan="2">Audio. There are also others</td>

possibilities, however. In particular, it should be noted that the example where the time structure analyzer 24 merely uses a value of the at least one time structure measurement 26 per instant according to the examples presented here, it is simply an embodiment and there are alternatives of according to which the analyzer of

IX / a / 2017/001240 temporal structure determines the measurement of temporal structure in a spectrally discriminating manner in order to obtain a value of the at least single measurement of temporal structure per spectral band of a plurality of spectral bands. Accordingly, the temporal structure analyzer 24 will then provide the controller 28 with more than one value of the at least one measurement of temporal structure 26 for the current frame 34a determined within the temporal region 36, namely one for said spectral band, while the spectral bands divide, for example, the general spectral range of the spectrogram 32.

Figure 7 illustrates the apparatus 10 and its use in an audio codec that supports the harmonic filter tool 30 according to the harmonic pre / post filter approach. Figure 7 illustrates a transform-based encoder 70 as well as a transform-based decoder 72 with the encoder 70 encoding the audio signal 12 in a data stream 74 and the decoder 72 receives the data stream to reconstruct the audio signal in the spectral domain as illustrated in 7 6 or, optionally, in the time domain illustrated in 78. It should be clear that the encoder and decoder 70 and 72 are distinct / separate entities and are jointly illustrated in Figure 7 for illustrative purposes only.

The transform-based encoder 70 comprises a transformer 80 that subjects the audio signal 12 to a transform. Transformer 80 can use a lapped transform as a critically sampled lapped transform, for example MDCT. In the example of Figure 7, the transform-based audio encoder 70 further comprises a spectral shaper 82 that spectrally shapes the signal of the audio spectrum as output of the transformer 80. The spectral shaper 82 can spectrally shape the spectrum of the audio signal according to a transfer function that is substantially inverse to a spectral perceptual function. The spectral perceptual function can be derived by linear prediction and thus, the information relative to the spectral perceptual function can be transferred to the decoder 72 within the data stream 74 in the form of, for example, linear prediction coefficients in the form of, for example, a pair

IX / a / 2017/001240 spectral of quantized lines of linear spectral frequency values. Alternatively, it is possible to use a perceptual model to determine the spectral perceptual function in the form of scale factors, one scale factor per band of scale factors, the bands can, for example, coincide with the Bark bands. The encoder 70 further comprises a quantizer 84 that quantizes the spectrum with, for example, a quantization function that is the same for all spectral lines. The spectrally formed and quantized spectrum is transported within the data stream 74 to the decoder 72.

In honor of completeness only, it should be noted that the order between transformer 80 and spectral shaper 82 has been selected in Figure 7 for illustrative purposes only. Theoretically, the spectral shaper 82 could generate the spectral formation within the time domain, that is, above the transformer 80. In addition, in order to determine the spectral perceptual function, the spectral shaper 82 could access the audio signal 12 in the time domain although it is not specifically indicated in Figure 7. On the decoder side, the decoder 72 is illustrated. in Figure 7 as comprising the spectral shaper 86 configured to form the internal spectrum

IX / a / 2017/001240 spectrally formed and quantized obtained from the data flow 74 with the inverse of the transfer function of the spectral shaper 82, i.e. substantially with the spectral perceptual function, followed by an optional inverse transformer 88. The inverse transformer 88 performs the inverse transformation relative to the transformer 80 and can, for example, perform a block-based inverse transformation followed by an overlap-sum process in order to cancel the aliasing of the domain of time, whereby the audio signal is reconstructed in the time domain.

As illustrated in Figure 7, a harmonic pre-filter may be comprised of the encoder 70 in a position above or below the transformer 80. For example, a harmonic pre-filter 90 above the transformer 80 may subjecting the audio signal 12 within the time domain to filtering to effectively attenuate the spectrum of the audio signal in the harmonics in addition to the transfer function or spectral shaper 82. Alternatively, the harmonic pre-filter can be arranged below the transformer 80 while this prefilter 92 performs or causes the same attenuation in the spectral domain. As illustrated in Figure 7, the

IX / a / 2017/001240 corresponding post-filters 94 and 96 are arranged inside decoder 72: the pre-filter 92, within the post-filter of the spectral domain 94 arranged above the reverse transformer 88 inversely forms the spectrum of the audio signal, inverse to the pre-filter transfer function 92, and in the case of using the pre-filter 90, the post filter 96 performs a filtering of the reconstructed audio signal in the time domain, below the reverse transformer 88, with a reverse transfer function to the transfer function of the pre-filter 90.

In the case of Figure 7, the apparatus 10 controls the harmonic filter tool of the audio codec implemented by the pair 90 and 96 or 92 and 94 by explicitly sending the control signals 98 via the audio codec of the data stream 74 next to the decoding to control the respective post-filter and, in line with the post-filter control on the decoding side, controlling the pre-filter on the encoder side.

In honor of completeness, Figure 8 illustrates the use of the apparatus 10 using an audio codec based on transformed elements 80, 82, 84, 86 and 88, however, the case in which the codec is illustrated here Audio supports only harmonic post-filter. In this case,

IX / a / 2017/001240 the harmonic filter tool 30 can be implemented by means of a post-filter 100 arranged above the reverse transformer 88 inside the decoder 72, in order to execute the post-harmonic filtering in the spectral domain, or by using a post-filter 102 disposed below the reverse transformer 88 to execute the harmonic post-filtering within the decoder 72 within the time domain. The mode of operation of post-filters 100 and 102 is substantially similar to that of post-filters 94 and 96: the purpose of these post-filters is to attenuate the quantization noise between harmonics. The apparatus 10 controls these post-filters by explicit signaling within the data stream 74, the explicit signaling is indicated in Figure 8 using reference 104.

IX / a / 2017/001240

<td>Is according</td><td colspan="2">has exposed the signal</td><td>from</td><td>control 98 or</td><td> 104</td><td>it is</td>

<td>sent by</td><td>example,</td><td>regularly,</td><td colspan="2">as per picture</td><td> 34 .</td><td>In</td>

<td>as for</td><td>picture,</td><td>it warns</td><td>what</td><td>the same</td><td>do not</td><td>They are</td>

<td>necessarily</td><td>of the</td><td colspan="2">same length</td><td>The length</td><td>from</td><td>the</td>

Pictures 34 may also vary.

The preceding description, especially that of the

Figures 2 and 3, reveal possibilities about how controller 28 controls the harmonic filter tool. As it arises from this exposure, it may be that the at least one time structure measurement measures an average or maximum energy variation of the audio signal within the time region 36. In addition, the controller 28 may include, within its control options, deactivation of the harmonic filter tool 30. This is illustrated in Figure 9. Figure 9 illustrates the controller 28 comprising a logic 120 configured to determine if a predetermined condition is satisfied by the at least one time structure measurement and the harmonicity measurement, to obtain a result of the check 122, which is of a primary nature and indicates whether the default condition is met or not. Controller 28 comprises a switch 124 configured to activate and deactivate the harmonic filter tool according to test result 122. If the result of check 122 indicates that the predetermined condition has been approved by logic 120, switch 124 indicates directly the situation by means of the control signal 14, or the switch 124 indicates the situation together with a degree of filter gain for the harmonic filter tool 30. That is, in the latter case, the switch 124 would not alternate between deactivating the harmonic filter tool 30 completely and turning on the harmonic filter tool 30

IX / a / 2017/001240 completely, only, but would put the harmonic filter tool 30 in some intermediate state that varies the intensity of the filtrate or filter gain, respectively. In that case, if the switch 124 also adapts / controls the harmonic filter tool 30 at some point between the complete activation / deactivation of the tool 30, the switch 124 may be based on the temporal structure measurement 26 and the harmonicity measurement 22 in order to establish the intermediate states of the control signal 14, that is, in order to adapt the tool 30. In other words, the switch 124 could determine the gain factor or adaptation factor to control the harmonic filter tool 30 based on measurements 26 and 22. Alternatively, the switch 124 uses for all states of the signal of control 14 that does not indicate the off state of the harmonic filter tool 30, the audio signal 12 directly. If the result of check 122 indicates that a predetermined condition has not been satisfied, then control signal 14 indicates the deactivation of the

IX / a / 2017/001240 harmonic filter tool 30.

As it emerges from the preceding description of the

Figures 2 and 3, the predetermined condition can be met if the at least one time structure measurement is less than a first predetermined threshold and the harmonicity measurement is, for a current and / or previous frame, greater than a second threshold. There may be another alternative: the predetermined condition can also be satisfied if the harmonicity measurement is, for a current frame, higher than a third threshold and the harmonicity measurement is, for a current frame and / or a previous frame, greater than one fourth threshold that decreases with an increase in pitch delay.

In particular, in the example of Figures 2 and 3, there are virtually three alternatives for which the predetermined condition is met, the alternatives depend on the at least one measurement of temporal structure:

one. A measurement of temporal structure &lt;threshold and combined harmonicity for current and previous table&gt; second threshold;

two. A measurement of temporal structure &lt;third threshold and (harmonicity for current and previous table)&gt; fourth threshold;

3. (A measurement of temporal structure &lt;fifth threshold or all temp measurements &lt;thresholds) and harmonicity for

IX / a / 2017/001240 the current table&gt; sixth threshold.

Thus, Figure 2 and Figure 3 reveal the implementation of possible examples for logic 124.

As illustrated above with respect to Figures 1 to 3, it is possible that the apparatus 10 is not only used to control a harmonic filter tool of an audio codec. Instead, the apparatus 10 can form, together with a transient detection, a system capable of performing the control of the harmonic filter tool as well as the detection of transients. Figure 10 illustrates this possibility. Figure 10 illustrates a system 150 composed of the apparatus 10 and a transient detector 152, and although the apparatus 10 emits the control signal 14 as set forth, the transient detector 152 is configured to detect transients in the signal Audio

12. To this end, however, the transient detector 152 exploits an intermediate result that occurs within the apparatus 10: the transient detector 152 uses for its detection the energy samples 52 temporarily or, alternatively, by spectrum-temporarily sampling the energy of the Audio signal, however, optionally evaluated energy samples within a time region in addition to time region 36 as within current frame 34a, for example. Based on the energy samples, the detector

IX / a / 2017/001240 transient 152 executes the detection of transients and

<td>signal</td><td>the</td><td>transient</td><td colspan="2">detected by</td><td>a</td><td>signal</td><td>from</td>

<td>detection</td><td> 154</td><td>. If</td><td>of the</td><td>previous example,</td><td>the</td><td>signal</td><td>from</td>

<td>detection</td><td>from</td><td colspan="2">transient</td><td>substantially</td><td colspan="2">indicates</td><td>the</td>

IX / a / 2017/001240 positions where the condition of equation 4 is met, that is, when the change in energy of the temporarily consecutive energy samples exceeds a certain threshold.

As is clear from the preceding discussion, a transform-based encoder such as that illustrated in Figure 8 or a drive-encoded excitation encoder may comprise or use the system of Figure 10 to switch a transform block and / or overlap length according to transient detection signal 154. In addition or alternatively, an audio encoder comprising or using the system of Figure 10 may be of the switching type. For example, USAC and EVS use switching between modes. Thus, this encoder could be configured to support switching between an excitation mode encoded by transform and a linear prediction mode excited by code and the encoder could be configured to perform the switching according to the transient detection signal 154 of the system of Figure 10. As for the excitation mode encoded by transform, the switching of the transform block and / or overlap length could, again, depend on the transient detection signal 154.

IX / a / 2017/001240

Examples of the benefits of previous embodiments

Example 1:

The size of the region where the temporary measurements for the LTP decision are calculated depends on the pitch (see equation (8)) and this region is different from the region where the temporary measurements for the transform length are calculated (usually current table and later).

In the example of Figure 11 the transient is within the region where the temporary measurements are calculated and thus the LTP decision is influenced. The motivation, as has been said, that an LTP for the current table, which uses the past samples of the indicated pitch delay segment, would reach a portion of the transient.

In the example in Figure 12 the transient is outside the region where temporary measurements are calculated and therefore does not influence the LTP decision. This is reasonable given that, unlike the previous figure, an LTP for the current picture would not reach the transitory one.

In both examples (Figure 11 and Figure 12) the transform length setting is decided on the temporary measurements only within the current frame, that is, the region marked with frame length. This means that in both examples, transient would not be detected in the current frame and preferably a single long transform (instead of many short ones) would be used.

Example 2:

In this case we describe the behavior of the LTP for the impulse and passage transients within the harmonic signal, as illustrated in the spectrogram of the Figure

13.

When the signal is encoded, it includes the LTP for the complete signal (since the LTP decision is based solely on the pitch gain), the output spectrogram looks as shown in Figure 14.

The waveform of the signal, whose spectrogram is in Figure 14, is presented in Figure 15. Figure 15 also includes the same signal filtered by low pass (LP) high pass (HP). In the signal filtered by LP the harmonic structure becomes clearer and in the signal filtered by HO the

IX / a / 2017/001240 location of the impulse as transitory and its trace is more evident. The level of the complete signal, the LP and HP signal is modified in the figure in honor of the presentation.

For short pulse transients (such as the first transient in Figure 13), long-term prediction produces transient repetitions as can be seen in Figure 14 and Figure 15. The use of long-term prediction during long pass-type transients (such as the second transient in Figure 13) does not introduce additional distortions since the transient is strong enough for a prolonged period and thus masks (simultaneously and post-masking) the portions of the signal constructed using this prediction. The decision mechanism allows LTP for step-type transients (to exploit the benefit of prediction) and disables LTP for the short-pulse transient (to prevent artifacts).

The energies of the segments computed in the transient detector are illustrated in Figure 16 and Figure 17. Figure 16 illustrates the impulse transient, Figure 17 illustrates the step transient. For the impulse transient in Figure 16 the temporal characteristics are calculated on the signal containing the current frame (N_new

IX / a / 2017/001240 segments) and the frame passed until the pitch delay (N_pastE (i) segments), since the TDVmax / ratio is above the ^ANDTD Vmin) threshold (—-—). For the transient type step in Figure 17, £ 0.375 G) and the ratio roVmax /_{they are}tra below the threshold ()

AND_TD Omin) 0-375 and thus only the energies of segments -8, -7 and -6 are used in the calculation of temporal measurements. These choices other than the segments where the temporary measurements are calculated lead to the determination of much higher energy fluctuations for the impulse transients and consequently to deactivate the LTP for the impulse transients and activate the LTP for the transient type transients.

Example 3:

However, in some cases the use of temporary measurements may not be convenient. The spectrogram in Figure 18 and the waveform in Figure 19 exhibit an extract of approximately 35 milliseconds since the beginning of Kalifornia by Fatboy Slim.

The LTP decision that depends on the Temporary Flatness Measurement and the Maximum Power Change disables the LTP for

IX / a / 2017/001240 this type of signal since it detects large temporary energy fluctuations.

This sample is an example of ambiguity among

IX / a / 2017/001240

<td colspan="2">transient</td><td>and the pulse train</td><td colspan="2">that form the</td><td colspan="2">bass signal</td>

<td>pitch</td><td></td><td></td><td></td><td></td><td></td><td></td>

<td>Such</td><td colspan="2">as can be seen</td><td>in</td><td>the figure</td><td> 20,</td><td>where he</td>

<td>abstract</td><td>from</td><td>600 milliseconds of</td><td>the</td><td>same signal</td><td>I know</td><td>presents,</td>

the signal contains a very short impulse transient

<td>repeated</td><td>(the</td><td colspan="2">spectrogram</td><td>I know</td><td colspan="2">produces using a</td><td>FFT</td><td>from</td>

<td>length</td><td colspan="2">short).</td><td></td><td></td><td></td><td></td><td></td><td></td>

<td>Such</td><td>how</td><td>may</td><td colspan="2">be observed</td><td>in the same</td><td>abstract</td><td>from</td><td> 600</td>

<td colspan="2">milliseconds</td><td>in the</td><td>Figure</td><td> 21</td><td>the signal</td><td colspan="2">It looks like</td><td>yes</td>

It contained a very harmonic signal with a low and changing pitch (the spectrogram is produced using a long length FFT).

This class of signals benefits from LTP since there is a clear repetitive structure (equivalent to the clear harmonic structure). Given the clear energy fluctuation (which can be seen in Figure 18, Figure 19 and Figure 20), the LTP would be deactivated by exceeding the threshold for the measurement of temporal flatness or for the maximum energy change. However, in our proposal, the LTP is activated because the normalized correlation exceeds the pitch delay dependent threshold (norm_corr (curr) &lt;= 1.2-T_int/ L).

Thus, the previous embodiments reveal, for example, a concept for a better harmonic filter decision for audio coding. It should be noted that it is possible to introduce certain changes from what has been revealed. In particular, as indicated, the audio signal 12 may be a voice or music signal and may be replaced by a pre-processed version of signal 12 for the purpose of pitch estimation, harmonicity measurement, or analysis or measurement of temporal structure. In addition, pitch estimation may not be limited to pitch delay measurements although, as those skilled in the art will warn, they may be performed by measurements of a fundamental frequency, in the time or spectral domain, which can easily become a delay. of equivalent pitch using an equation such as pitch delay = sampling frequency / pitch frequency. In general, the pitch estimator 16 estimates the pitch of the audio signal which, in turn, is manifested in the pitch delay and pitch frequency.

Despite having described some aspects in the

IX / a / 2017/001240 context of an apparatus, it is clear that these aspects represent a description of the corresponding method, where a block or device corresponds to a stage of the method or a characteristic of a stage of the method. Similarly, the aspects described in the context of a method step further represent a description of a corresponding block, element or characteristic of a given apparatus. Part or all stages of the method can be executed (or used) by means of an apparatus such as a microprocessor, a programmable computer or an electronic circuit. In some embodiments, some, one or more of the most important steps of the method may be executed by this apparatus.

The encoded audio signal of the invention can be stored in a digital storage medium or it can be transmitted by a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

According to certain implementation requirements, the embodiments of the invention can be implemented in hardware or software. The implementation can be executed using a digital storage medium, for example a floppy disk, a DVD, a Blu-Ray, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, with control signals electronically readable stored, cooperating

IX / a / 2017/001240 (or are able to cooperate) with a computer system program so that the respective method is executed. Consequently, the digital storage medium can be computer readable.

Some embodiments according to the invention comprise a data carrier with electronically readable control signals, which are capable of cooperating with a programmable computer system, in order to execute one of the methods described herein.

In general, the embodiments of the present invention can be implemented as a computer product with a program code, the program code can execute one of the methods when running on a computer. The program code may for example be stored on a machine-readable media.

Other embodiments include the computer program for executing one of the methods described, stored on a machine-readable media.

In other words, an embodiment of a method of the invention is, consequently, a computer program that has a code to execute one of the methods described, when the program runs on a computer.

IX / a / 2017/001240

A further embodiment of the methods of the invention is, consequently, a data carrier (or a digital storage medium, or a computer-readable medium) comprising, recorded therein, the computer program for executing one of the methods described. The data carrier, the digital storage medium or the recorded media are typically tangible and / or non-transient.

A further embodiment of the method of the invention is, consequently, a flow of data or a sequence of signals representing the computer program for executing one of the methods described herein. The data throughput or the sequence of signals can for example be configured to be transferred by a data communication connection, for example, via the Internet.

A further embodiment comprises a processing means, for example a computer, or a programmable logic device, configured for or adapted to execute one of the methods described herein.

A further embodiment comprises a computer that has the program installed to execute one of the methods described herein.

A further embodiment according to the invention comprises an apparatus or a system configured to transfer (for example, electronically or optically) a program

IX / a / 2017/001240 computer to execute one of the methods described herein to a receiver. The receiver may, for example, be a computer, a mobile device, a memory or the like. The apparatus or system may, for example, comprise a file server to transfer the computer program to the receiver.

In some embodiments, a programmable logic device (for example a set of programmable gates per field) can be used to execute some or all of the functionalities of the methods described herein. In some embodiments, a set of programmable gates per field may cooperate with a microprocessor in order to execute one of the methods described herein. In general, the methods are preferably executed by any apparatus.

The embodiments described above are merely illustrative of the principles of the present invention. It is understood that those skilled in the art will notice modifications and variants to what has been revealed. Accordingly, the invention will only be limited by the appended claims and not by specific details set forth for descriptive and illustrative purposes of the embodiments thereof.