Coding transmission equipment for carrying out coding with adaptive quantization.

BACKGROUND OF THE INVENTION

1. Field of the Invention -

The present invention relates to coding tranmission equipment in which an input signal, such as a voice signal or a picture signal, is coded and transmitted to a receiver side from a transmitter side and a received signal is decoded to become an original signal at the receiver side. In particular, the present invention relates to low bit rate coding equipment (bandwidth compression coding equipment), such as an ADPCM (Adaptive Differential Pulse Code Modulation) system CODEC (COder and DECoder), which carries out low bit rate coding of a voice or picture signal.

When the voice signal etc., is transmitted, bandwidth compression is required to ensure an efficient utilization of a transmission line. The low bit rate coding transmission equipment for the voice signal carries out information compression while maintaining the quality of the voice signal, which enables a reduction of line costs for the transmission of the voice information in equipment used for , for example, mobile radio communication, satellite communication, or office inter communication system, and a reduction of storage capacity needed, for example, for the storage of voice information in an audio response system.

2. Description of the Related Art

ADPCM coding equipment is related to the present invention in that it carries out coding by adaptive quantization. In the ADPCM coding equipment, a prediction error E corresponding to a difference between a current input signal X and a predicted value X obtained on the basis of past input signals is produced and quantized at the transmitter side. At this time, adaptive quantization, in which the size of a quantizing step is increased or decreased in accordance with a quantizing level of a signal sampled just prior to the current sampled signal, is carried out in order to realize a quick adaptation to a sudden change of the predicted error, instead of uniform quantization, whereby a high quality recovered signal is realized.

In this case, however, the sizes to which the quantizing step can increased or decreased are limited, and therefore, when the input signals are voice signals, for example, a voiced sound, an unvoiced sound, or a voicelessness, etc., the nature of the signals changes from moment to moment and the coding system of the prior art cannot cope with these rapid variations in the voice signal. As a result, an optimum quantization is not realized, and thus the voice recovered at the receiver side has a poor quality.

SUMMARY OF THE INVENTION

Accordingly, it is an object of the present invention to provide coding transmission equipment which can obtain a high quality recovered signal, for example, a voice signal recovered at a receiver side.

According to a fundamental aspect of the present invention, there is provided coding transmission equipment having a transmitter device and a receiver device wherein, the transmitter device comprises a plurality of adaptive coding units including an adaptive quantizer for outputting a quantized value according to an input signal, and an error calculating unit for calculating the quantization error of the adaptive quantizer, each of the adaptive coding units having different adaptive quantizer processing characteristics; an evaluating unit for deciding which is the optimum adaptive coding unit to carry out an optimum quantization, from among the plurality of adaptive coding units, at every frame of the input signal on the basis of the quantization errors input from the plurality of adaptive coding units respectively; and a sending unit for selecting the optimum quantized value of the optimum adaptive coding unit from among the quantized values input from the plurality of adaptive coding units, respectively, on the basis of the decision of the evaluating unit, and transmitting the optimum quantized value together with the decision information; the receiver device comprises a dividing unit for dividing a signal received from the transmitter device into the optimum quantized value and the decision information; an adaptive decoding unit for decoding the optimum quantized value obtained by the dividing unit into the original input signal; and an optimum controlling unit for controlling the processing characteristic of the adaptive decoding unit to an optimum processing characteristic according to the decision information obtained by the dividing unit.

According to another aspect of the present invention, the transmitter device further comprises a parameter copy unit in which parameters for determining an inner condition of the adaptive coding unit are input from all of the adaptive coding units, and an optimum parameter of the optimum adaptive coding unit is selected and sent to other adaptive coding units at every frame according to the decision of the evaluating unit; wherein each of the other adaptive coding units starts the signal process for a next frame after copying the optimum parameter sent from the parameter copy unit, as an own parameter.

According to still another aspect of the present invention, the transmitter device further comprises an error correction coding unit for carrying out an error correction coding process for the decision information from the evaluating unit, whereby the decision information processed by the error correction coding is sent to the receiver device; and the receiver device further comprises an error correcting unit for carrying out the error correcting process for the decision information obtained by the dividing unit, whereby the error corrected decision information is sent to the optimum controlling unit.

According to yet another aspect of the present invention, the receiver device further comprises a post-processing filter disposed at a latter stage of the adaptive decoding unit, wherein the parameter of the post-processing filter is varied according to the parameter of the adaptive decoding unit.

BRIEF DESCRIPTION OF THE DRAWINGS

Examples of coding transmission equipment in accordance with the present invention will be described with reference to the accompanying drawings, in which <ul id="ul0001" list-style="none"><li>Fig. 1 shows a prior art ADPCM coder;</li><li>Fig. 2 shows a prior art ADPCM decoder;</li><li>Fig. 3 shows an embodiment of an ADPCM coder at a transmitter side of the coding transmission equipment according to the present invention;</li><li>Fig. 4 shows an embodiment of an ADPCM decoder at a receiver side of the coding transmission equipment according to the present invention;</li><li>Fig. 5 shows an adaptive differential coding unit of the ADPCM coder of Fig. 3 in more detail;</li><li>Fig. 6 shows another embodiment of the ADPCM coder according to the present invention;</li><li>Fig. 7 shows a further embodiment of the coding transmission equipment according to the present invention;</li><li>Fig. 8 shows an example of an error correcting code;</li><li>Fig. 9 shows a format of the transmission signal including the error correcting code;</li><li>Fig. 10 shows an example of an error correction at the receiver side;</li><li>Fig. 11 shows still another embodiment of the ADPCM decoder at the receiver side according to the present invention; and</li><li>Fig. 12 is a view for explaining a post-processing filter in detail.</li></ul>

DESCRIPTION OF THE PREFERRED EMBODIMENTS

For a better understanding of the preferred embodiment, the problems in the prior art will be first explained with reference to Figs. 1 and 2.

Figure 1 shows an adaptive differential coder on the transmitter side in a prior art low bit rate coding transmission equipment for a voice signal. In Fig. 1, the coder comprises a subtractor 10 for calculating a prediction error E(n) corresponding to a difference between an input voice signal X(n) and a predicted value X(n) to be output; an adaptive quantizer 11 for adaptive quantizing of the prediction error E(n) to output the quantized value I(n) as a transmission signal; an adaptive inverse quantizer 12 for adaptive inverse quantizing of the quantized value I(n); an adder 13 A for adding the inverse quantized value E(n) of the prediction error E(n) from the inverse quantizer 12 and A the predicted value X(n); an adaptive zero predictor 14 for producing a zero predicted value X̂(n)_z; an adaptive pole predictor 15 for producing a pole predicted value X̂(n)_p; and an adder 16 for adding the zero predicted value X̂(n) and the pole predicted value X̂(n) to produce the predicted value X̂(n).

In this adaptive differential coder, the quantizer 11, the inverse quantizer 12, the zero predictor 14, and the pole predictor 15 are adaptive- controlled. For example, in the quantizer 11 and the inverse quantizer 12, the adaptive quantization wherein the quantizing step size Δ(n) is updated according to the following formula is carried out. Δ(n+1) = Δ(n)^γ x M(I(n)) ... (1) where, y is a coefficient for gradually reducing the influence of the line error, which has a value close to and lower than 1; and M(I(n)) is a step size updating coefficient (or a step size updating speed).

When the quantized value I(n) from the quantizer 11 is, for example, a 1 bit output signal, the updating coefficient M(I(n)) assumes a value a close to and lower than 1, for example, a = 0.93, when the quantized value I(n) is "0", and a value β close to and higher than 1, for example, β = 1.31, when the quantized value I(n) is "I". Also, when the quantized value I(n) is a plurality bit output signal, the step size updating coefficient M(I(n)) is increased in proportion to the increment of the number indicated by the plurality bit output signal.

Figure 2 shows an adaptive differential decoder on the receiver side in a prior art low bit rate coding transmission equipment.

In Fig. 2, the decoder comprises an inverse quantizer 71', an adder 73, a zero predictor 75, a pole predictor 76, and an adder 77.

In this decoder, the quantized value I(n) of the prediction error transmitted from the transmitter side is inverse-quantized by the inverse quantizer 71 to recover the prediction error, and the zero predicted value X̂(n)_z and the pole predicted valueX̂(n)_p are produced by the zero predictor 75 and the pole predictor 76 on the basis of the recovered past prediction errors respectively. Then, the predicted value X̂(n) obtained by adding this zero predicted value X̂(n)z and the pole predicted value X̂(n)p by the adder 73 is added to the inverse-quantized prediction error from the inverse quantizer 71' by the adder 73, to recover the voice signal S(n). The processing characteristics of the inverse quantizer 71', the zero predictor 75, and the pole predictor 76 at the receiver side are the same as that of the inverse quantizer 12, the zero predictor 14, and the pole predictor 15 at the transmitter side, respectively.

In the differential coder of Fig. 1, the step size updating coefficient M(I(n)) of the quantizer 11 and the inverse quantizer 12 is a fixed value, since the values a and 3 which determine the updating coefficient M(I(n)) are fixed. Therefore, sometimes the optimum quantization is not carried out at the quantizer 11, since updating coefficient M(I(n)) of the step size Δ(n), which is a threshold for quantizing the prediction error E(n), may not sufficiently follow the change of the prediction error E(n).

In particular, when the input signal is a voice signal, the nature of the signal, for example, a voiced sound, unvoiced sound, or silence, changes from moment to moment and frequently varies, and therefore, the optimum quantization is not carried out since the fixed step size updating coefficient M(I(n)) cannot follow the change of the voice signal.

A preferred embodiment of the present invention will now be explained with reference to Figs. 3 to 5. Figures 3 to 5 show a embodiment of the coding transmission equipment, according to the present invention, which is adapted to transmission equipment using the ADPCM coding system. Figure 3 shows an ADPCM coder at the transmitter side, and Fig. 4 shows an ADPCM decoder at the receiver side.

In Fig. 3, 1₁ to 1_k are the k numbers of adaptive differential coding units in which the step size updating coefficients M(I(n)), to M(I(n)), of quantizers (in more detail, values a and β described above) are different, respectively. The fundamental structure is the same as the adaptive differential coder shown in Fig. 1, except that a subtractor is provided which calculates and outputs a quantization error e(n) corresponding to a differential value between the prediction A error E(n) and the inverse-quantized value E(n) of the prediction error E(n) from the inverse quantizer 12.

Figure 5 shows a more detailed structure of one of the adaptive differential coding units l_l to l_k. In Fig. 5, the subtractor 10 calculates the prediction error E(n) corresponding to the differential value between the input voice singal X(n) and the predicted value X(n) and sends this prediction error E(n) to the adaptive quantizer 11. The adaptive quantizer 11 quantizes the prediction error E(n) and outputs a quantized value I(n). This quantized value I(n) is introduced to the inverse quantizer 12, which carries out the inverse quantization. At the same time, the quantized value I(n) is transmitted to the receiver A side. The inverse quantized value E(n) of the prediction error E(n) from the inverse quantizer 12 is introduced to the zero predictor 14 and the subtractor 17, and to the pole predictor 15 via the adder 13. The zero predictor 14 produces a zero predicted value X(n) z and the pole predictor 15 produces a pole predicted value X(n) . The zero predicted value X(n) and the pole predicted value X(n) are added by the adder 16 to produce a predicted value X(n), which is then sent to the subtractor 10. The subtractor 17 calculates the differential value between the prediction error E(n) and A the inverse quantized value E(n) thereof, and outputs the result as a quantization error e(n).

The quantizer 11 and the inverse quantizer 12 are adaptive controlled type devices, and the quantizing step size Δ(n) is updated by the quantizing step adapter (not shown) according to the formula (1) described previously. This quantizing step size Δ(n) is a parameter for deciding inner conditions of the quantizer 11 and the inverse quantizer 12.

The zero predictor 14 and the pole predictor 15 are also adaptive control type devices, and are constituted by circuits comprising taps (time delay elements) and coefficient multipliers, having the orders t and m, respectively. The zero predictor 14 and the pole predictor 15 hold prediction coefficients updated in sequence, and from the viewpoint of the transfer function as a filter, hold tap data containing past records.

That is, in the zero predictor 14, zero prediction coefficients C_z(1, n) ~ C_z(ℓ, n) of each coefficient multiplier are updated by the prediction coefficient adapter (not shown) according to the following formula, respectively.<maths id="math0001"><img file="EP0206352A2_D0001.tif" /></maths>where, L_zand D are constant, and sgn( ) is a signature function which indicates a signature of "+" or "-".

The zero prediction coefficients C (1, n) to C_z(ℓ, n) and tap datas E(n-L) to E(n-ℓ) stored at each tap are parameters for determining the inner condition of the zero predictor 14.

Also, in the pole predictor 15, pole prediction coefficients C_p (ℓ, n) to Cp(m, n) of each coefficient multiplier are updated by the prediction coefficient adapter (not shown) according to the following formula, respectively.<maths id="math0002"><img file="EP0206352A2_D0002.tif" /></maths>where, L_p and D_p are constant.

The pole prediction coefficients C (1, n) to C_p (m, n) and tap data S(n-1) to S(n-m) stored at each tap are the parameters for determining the inner condition of the pole predictor 15.

In Fig. 3, each component of Fig. 5 is represented in such a manner that the quantizer 11 is represented by Q_l to Q_k , the inverse quantizer 12 by Q₁^-1 to Q_k^-1, the zero predictor 14 by H_zl to H_zk , and the pole predictor 15 by H_pl to H_pk , respectively.

In Fig. 3 quantization errors e(n)₁ to e(n)_k are introduced to an evaluating unit 2. The evaluating unit 2 comprises a quantization error power calculating unit 21 and an optimum quantization deciding unit 22. The calculating unit 21 calculates powers of the quantization errors e(n)₁ to e(n)_k , sample by sample respectively, during one frame of the voice signal X(n) (for example, 16 samples), and integrates them, respectively, at each frame, then outputs the summed powers of each of the quantization errors e(n) to e(n)k at the end of the frame to the deciding unit 22. The deciding unit 22 decides the optimum coding unit 1 opt, which carries out the optimum quantization, among the adaptive differential coding units 1₁ to l_k , by selecting the minimum calculated quantization error power from that of the quantization errors e(n)1 to e(n)_k . Then, the deciding unit 22 outputs an optimum coder number signal Nopt, indicating the number of the optimum coding unit 1 opt decided as aforementioned, to a selector 3 and a multiplexer 4.

The quantized values I(n)₁ to I(n)_k from adaptive differential coding units 1₁ to l_k are introduced to the selector 3, respectively. This selector is constituted by a memory such as a RAM, which can store quantized values I(n)₁ to I(n)k of one frame. When the coder number signal Nopt is received at the end of one frame, the selector 3 sends one frame of the optimum quantized value I(n)_opt of the optimum coding unit 1 opt, which is selected by using the coder number signal Nopt, as an address to the multiplexer 4. The multiplexer 4 adds the coder number signal Nopt from the evaluating unit to the optimum quantized value I(n)opt, and transmits a multiplexed signal to the receiver side after converting the signal to a transmission line code.

In the receiver of Fig 4, 6 is a demultiplexer, and 7 a decoder. The decoder 7 comprises an adaptive inverse quantizer 71; a controller 72 capable of controlling the step size updating coefficient M(I(n)) of the inverse quantizer 71; an adaptive predictor 74 constituted by a zero predictor 75, a pole predictor 76, and an adder 77; and an adder 73.

The operation of the above described coding transmission equipment is explained hereinafter.

The voice signal X(n) is input to each of adaptive differential coding units l_l to l_k in parallel, an adaptive differential coding process is carried out at each unit, and the quantized values I(n)_l to I(n), and the quantization errors e(n)_l to e(n)_k are output. The quantization error power calculating unit 21 calculates the power of the input quantization errors e(n)_l , to e(n)_k frame by frame, respectively, then sends the calculated results to the optimum quantization deciding unit 22. The deciding unit 22 decides the optimum coding unit 1 opt for outputting the quantization error e(n), having a minimum quantization error power, which carries out the optimum quantization and then sends the coder number signal Nopt indicating the coder number of the decided optimum coding unit 1 opt to the selector 3 and the multiplexer 4.

In response to this coder number signal Nopt, the selector 3 sends one frame of the optimum quantized value I(n)opt of the optimum coding unit lcpt indicated by the coder number signal Nopt to the multiplexer 4. The multiplexer 4 adds the coder number signal Nopt to the optimum quantized value I(n)opt and transmits the multiplexed signal to the receiver side.

In the receiver side, the received signal is divided into the quantized value I(n)opt of the prediction error E(n) and the coder number signal Nopt, and input to the inverse quantizer 71 and the controller 72, respectively. The inverse quantizer 71 inverse-quantizes the quantized value I(n)opt to obtain the original prediction error E(n), then sends the prediction error E(n) to the adder 73. The adder 73 adds the inverse-quantized prediction error E(n) and the predicted value X(n) predicted by the predictor 74 on the basis of past recovered voice signals to recover the current voice signal S(n). The predictor 74 carries out the adaptive prediction according to the output signal of the inverse quantizer 71.

On the other hand, the controller 72 controls the step size updating coefficient M(I(n)) of the inverse quantizer 71 to coincide with the step size updating coefficient M(I(n))opt of the optimum coding unit 1 opt selected as the coding unit carrying out the optimum quantization for a relevant frame at the transmitter side. That is, the controller 72 reads out the step size updating coefficient M(I(n)) from a ROM by using the coder number signal Nopt as an address, then sends it to the inverse quantizer 71, and an inverse quantization step adapter in the inverse quantizer 71 changes the updating coefficient M(I(n)).

As described above, in the coding transmission equipment of the present invention, the coder of the transmitter side selects the optimum coding unit which carries out the optimum quantization from among a plurality of adaptive differential coding units having different step size updating coefficients, then transmits the quantized value thereof with the coding unit number, and the decoder of the receiver side recovers the original voice signal from the received quantized value by using the optimum step size updating coefficient indicated by the received coding unit number, whereby an improvement of the quantity of the recovered voice signal is achieved.

Although a preferred embodiment has been described heretofore, various modifications and alterations are possible within the scope of the present invention.

For example, means for changing the optimum updating coefficient M(I(n)) in accordance with the received coder number signal Nopt at the receiver side is not limited to the above described example. The receiver can have a plurality of inverse quantizers having different step size updating coefficients corresponding to those of the transmitter side, and can recover the voice signal by selecting the optimum inverse quantizer by the selector according to the received coder number signal. Also the receiver can have a plurality of decoders having difference processing characteristics corresponding to each coding unit at the transmitter side, and can select the optimum decoder according to the received coder number signal.

Figure 6 shows an other embodiment of the coder according to the present invention. This coder is improved so that a better quality recovered voice signal is obtained at the receiver side in comparison with the coder of Fi^g 3.

The quantizer, the inverse quantizer, the zero predictor, and the pole predictor in each adaptive differential coder l_l to l_k are adaptive controlled, and therefore, the parameters deciding the inner condition of the coding unit, for example, the quantization step size a(n) of the quantizer and the inverse quantizer and the prediction coefficient and tap data of the predictor, etc., vary successively at each of the differential coding units 1₁ to l_k according to the processing procedure of the input signal, and become different at each differential coding unit.

Accordingly, when the prior selected differential coding unit is changed to the other differential coding unit currently selected at a boundary between frames, in order to select the optimum differential coding unit, the aforementioned parameter of the prior selected coding unit and that of the currently selected coding unit are different. As a result, when the differential coding unit is switched, a discontinuity in the process of the input signal arises at the boundary between frames due to the difference between these parameters, therefore the discontinuity of the recovered signal also arises at the boundary between frames in the receiver side, and thus the recovered voice has a poor quality. The coder shown in Fig. 6 is intended to solve this problem.

The difference between the coder of Fig. 6 and the coder of Fig. 3 is that the coder of Fig. 6 is provided with a parameter copy controller 6. That is, parameters P1 to _Pk from the differential coding units 1₁' to l_k', which parameters determine the inner condition of the coding unit, and the coder number signal Nopt from the evaluating unit 2 are input to the-parameter copy controller 6. Here, parameters P_l to P_K are quantizing step sizes Δ(n)_l to Δ(n)_k , zero prediction coefficients C_zl to C_zk , pole prediction coefficients C_pl to C_pk , tap datas E₁ to E_k, and tap datas S1 to S_k. The parameter copy controller 5 selects the parameter Popt of the optimum coding unit lopt according to the coder number signal Nopt and then sends it to all of the differential coding units l_l' to l_k'.

The operation of the coder shown in Fig. 6 will be explained below.

The optimum coding unit lopt is decided by the evaluating unit at the end of one frame of the voice signal X(n), and then the coder number signal Nopt is sent to the selector 3, the multiplexer 4, and the parameter copy controller 5. The selector 3 sends one frame of the quantized value I(n)opt of the optimum coding unit lopt indicated by the signal Nopt to the multiplexer 4, and the multiplexer 4 then transmits this optimum quantized value I(n)opt with the coder number signal Nopt to the receiver side.

On the other hand, the parameter copy controller 5 selects the optimum parameter Popt of the optimum coding unit 1 opt corresponding to the coder number signal Nopt from among the input parameters P_l to P_k , and then sends that parameter Popt to all of the differential coding units l_l' to l_k'. Accordingly, all of the adaptive differential coding units copy the received optimum parameter Popt as an own parameter at the end of one frame.

That is, at each adaptive differential coding unit, the quantizing step size Δ(n)opt is written in the quantizer 11 and the inverse quantizer 12, the zero prediction coefficient Cz.opt and the tap data Eopt are written in the coefficient multiplier of the zero predictor 14 and the tap thereof, respectively, and the pole prediction coefficient Cp.opt and the ta^p data So^pt are written in the coefficient multiplier of the pole predictor 15 and the tap thereof, respectively.

As a result, all of the adaptive differential coding units l_l' to l_k' have the same inner condition as that of the currently selected adaptive differential coding unit 1' opt at the time of starting the next frame. Accordingly, when, for example, the adaptive differential coding unit 1,' is selected as the optimum coding unit at current frame and then the adaptive differential coding unit 1₂' is selected at next frame, the parameter P_l of the differential coding unit 1₁' at the end point of the current frame coincides with the parameter P₂ of the differential coding unit 1₂' at the starting point of the next frame, and therefore, no discontinuity arises at the boundary between the current and next frames.

Figure 7 shows a further embodiment of the present invention. In the coding transmission equipment shown in Figs. 3 to 6, if the coder number signal Nopt indicating the optimum coding unit number is erroneous, the influence of the error extends over several frames, and thus the recovered voice at the receiver side has a poor quality. The equipment of Fig. 7 is intended to solve this problem.

That is, the transmission equipment of Fig. 7 is provided with an error correction circuit for correcting the error arising in the transmitted optimum coder number signal Nopt.

In Fig. 7, the coder number signal from the evaluating unit 2 is input to the multiplexer 4 via an error correcting coder 81, and then transmitted to the receiver side. On the other hand, at the receiver side, the coder number signal Nopt divided by the demultiplexer 6 is input, via an error correcting decoder 82, to the decoder 7, wherein the signal Nopt is input to the controller 72 as shown in Fig. 4 for controlling the updating coefficient M(I(n)) of the inverse quantizer. The error correcting coder 81 and the error correcting decoder 82 carry out the error correction as follows.

First, in this example, the number of differential coding units is four, and each coding unit is given the coder numbers "00", "01", "lC", and "11". Figure 8 shows an example of an error coding code, Fig. 9 shows a format of one frame of the transmission signal, and Fig. 10 shows an example of an error correction at the receiver side.

As shown in Fig. 8 the error correcting code ECC is obtained by converting "0" and "1" of each digit of the coder number into'"00" and "11" respectively. The transmission code is configured by arranging the high order digit error correcting code ECC(h) of the coder number, the high order digit N(h) thereof, the low order digit error correcting code ECC(ℓ) the coder number, and the low order digit N(t) thereof, in sequence. Therefore, the coder number constituted by two digits is converted into the transmission code constituted by six digits and transmitted to the receiver side.

In the receiver side, the error correcting decoder 82 decides whether each of the high and low order digits of the received transmission code is "0" or "1" by a majority of "0" and "1" thereof, and corrects all digits of each of the high and low order digit of the received transmission code to "000" or "111" according to the majority decision, thereby recovering the correct high and low order of the digits of the coder number.

For example, Fig. 10 shows the error correction of high order digits of the received transmission code at the receiver side. As shown in Fig. 10, "001", "010", and "100" of the high order digits wherein "1" may be erroneous are corrected to "000". Also, "0ll", "101", and "110" wherein "0" may be erroneous are corrected to "lll".

Various kinds of error correcting codes are known, and a suitable error correcting code may be adopted according to the number of differential coding units and kind of coder number to be transmitted, etc.

Assuming that a sampling frequency of the input voice signal is 6.4 kHz, and the quantized value code of the prediction error is 2 bits at each sample point, an example of a bit allocation of the frame which transmits the quantized value codes of 16 sample points will be explained below.

One frame is constituted by 39 bits wherein 32 bits are the quantized value codes, each of which is 2 bits of 16 samples, 2 bits are the coder number, and 4 bits are error correcting codes ECC(h) &amp; ECC(t). Each frame is transmitted during 2.5 ms (= 1/400 sec) corresponding to a 16 sample period at a sampling frequency of 6.4 KHz, therefore the transmission rate is 15.6 Kbit/sec.

Figure 11 shows still another embodiment of the receiver side of the coding transmission equipment according to the present invention. The difference between the embodiment of Fig. 5 and this embodiment of Fig. 11 is that here a post-processing filter is provided. The post-processing filter carries out the signal process of the decoded signal in order to improve an auditory sense of the recovered voice. This post-processing filter is disclosed in a paper entitled "Enhancement of ADPCM Speech by Adaptive Post-filtering" by V. Ramamoorthy &amp; N.S. Jayant, published at the "ICC '85" held in the USA. According to this paper, when a transfer function A(z) of the pole predictor 76 of the predictor 74 and a transfer function B(z) of the zero predictor 75 thereof are as follows,<maths id="math0003"><img file="EP0206352A2_D0003.tif" /></maths><maths id="math0004"><img file="EP0206352A2_D0004.tif" /></maths>a transfer function A' (z) of a pole post-processing filter 86 of the post-processing filter 8 and a transfer function B'(z) of a zero post-processing filter 85 is assumed to be as follows.<maths id="math0005"><img file="EP0206352A2_D0005.tif" /></maths><maths id="math0006"><img file="EP0206352A2_D0006.tif" /></maths>(0≦ α ≦ 1, 0 ≦ β ≦ 1)

The principal structure of the predictor 74 and the post-processing filter 8 disclosed in this paper is shown in Fig. 12. In Fig. 12, the reference symbols of the coefficient multipliers denote the coefficients thereof, respectively, for simplification of the explanation.

The predictor of the decoder 7 shown at left side of Fig. 12 comprises the zero predictor 75 including six time delay elements Z^-1, six coefficient multipliers b_l to b₆, and an adder Al adding output signals of multipliers b₁ to b₆ , which realizes the transfer function B(z), and the pole predictor 76 comprising two time delay elements Z^-1, two coefficient multipliers a₁ and a₂ , and an adder A2 adding output signals of the multipliers a₁ and a₂, which realizes the transfer function A(z). Output signals of the zero predictor 75 and the pole predictor 76 are added to the signal E(n) obtained by inverse quantizing the quantized value from the transmitter side at the inverse quantizer 71 (Fig. ll) by the adder 73, to obtain the recovered voice signal S(n).

A block at the right side in Fig. 12 is a post-processing filter 8', which comprises a zero post-processing filter 85, a pole post-processing filter 86, and an adder 83. The zero post-processing filter 85 is further provided with 3 coefficient multipliers β and β² to β⁶ inserted between coefficient multipliers b_l to b₂ and the adder Al, respectively, in comparison with the zero predictor 75 in order to carry out the operation of β coefficient which is a power of β.

The pole post-processing filter is provided with a coefficient multiplier a and α² inserted between coefficient multipliers a₁ and a₂ and the adder A2, respectively, in comparison with the pole predictor 76. Output signals of post-processing filters 85 and 86 are added to the recovered voice output Ŝ(n) at the adder 83, whereby the auditory sense of the recovered output is improved, and then output as an improved output Z.

The aforementioned coefficients b₁ to b₆ , a₁ and a₂ are updated to adapt to the input signal, and must have the same values at the predictor and the post-processing filter. Also, the coefficients a and β at the post-processing filter are updated to adapt to the input signal.

In Fig. 11, the post-processing filter 8, such as the filter shown in Fig. 12, is disposed at the output side of the decoder 7. This post-processing filter 8 is provided with a coefficient weighting circuit 88 for controlling the coefficients a and β of the zero and pole post-processing filters 85 and 86.

The coefficients of the coefficients multipliers b₁ to b₆ of the zero post-processing filter 85 and the coefficient multipliers a₁ and a₂ of the pole post-processing filter 86 are kept at the same values as those of the coefficients b₁ to b₆ of the zero predictor 75' and coefficients a₁ and a₂ of the pole predictor 76', respectively, by the coefficient weighting circuit 88. Also, the power coefficient β of the zero post-processing filter 85 and the power coefficient a of the pole post-processing filter 86 are adjusted to suitable values for carrying out a suitable post-processing by the coefficient weighting circuit 88. The coefficients a and B are set to a value between 0 to 1. Note in this example, a is 1.0 and β is 0.2.

As described above, the recovered voice output having a further improved quality can be obtained by adding the post-processing filter having a characteristic which is changed according to the characteristic of the adaptive differential coding unit at the transmitter side.

In the embodiments explained hereinbefore, the present invention is applied to the ADPCM system. But, of course, the present invention is not limited to the ADPCM system, namely, the present invention may be adapted to systems wherein the adaptive quantization is carried out as, for example, APCM (Adaptive Pulse Code Modulation), ADM (Adaptive Delta Modulation), SBC (Sub Band Coding), VRC (Variable Rate Coding), APC (Adaptive Predictive Coding), and ATC (Adaptive Transform Coding), etc.