Method and system for processing sound signals in a vehicle multimedia system

Field of the Invention

0001This invention relates to a method and system for processing sound signals in a vehicle multimedia system.

Related Art

0002Modern vehicle multimedia systems often comprise vehicle interior communication (intercom) systems which can improve the communication between passengers, especially when high background noise levels are present. Particularly, it is important to provide means for improving the communication between passengers in the backseat and the front seat of the vehicle, since the direction of speech produced by a front passenger is opposite to the direction in which the passenger in the rear seat is located. To improve the communication, speech produced by a passenger is recorded with one or more microphones and reproduced by loudspeakers that are located in close proximity to the listening passengers. As a consequence, sound emitted by the loudspeakers may be detected by the microphones, leading to reverb/echo or feedback. The loudspeakers may also be used to reproduce audio signals from an audio source, such as a radio, a CD player, a navigation system and the like. Again, these audio signal components are detected by the microphone and are put out by the loudspeakers, again leading to reverb or feedback. Furthermore, the background noise level inside the vehicle constantly changes. In city traffic the background noise level will be very different when the vehicle is standing, e.g. in front of a red traffic light, or when the vehicle is driving at moderate speed. At high vehicle speeds the change in background noise is even higher, for example when accelerating from 70 km/h to 130 km/h. In order to still enable intervehicle communication or to obtain audible sound from the multimedia system, the vehicle passengers have to continuously adjust the volume/amplification of the multimedia system. Adjustment of the volume can distract the driver and can lead to dangerous situations.

0003Furthermore, the vehicle passengers may want to be entertained during their journey. For this purpose, a karaoke system can be provided inside the vehicle. Such a karaoke system suffers from the same drawbacks as a vehicle intercom system, meaning that the reproduction of the voice from a singing passenger is prone to reverb and feedback. The volume of the sound output of the karaoke system also needs to be constantly adjusted by the passengers, depending on the background noise level inside the passenger compartment.

0004In the art systems are known that use several microphones to record a voice signal produced by a speaking passenger. The detected sound signals are then processed for beamforming to obtain a directional sound signal in which sound originating from a direction other than a preferred direction is attenuated. The preferred direction is generally the direction in which the speaking passenger is located. That way sound originating from other directions like sound emitted from a loudspeaker is attenuated in the beam formed detected signal, whereby reverberation and feedback is reduced. Furthermore, background noise originating from directions other than the speaking passenger direction is attenuated by the beam former. Such beamforming systems are known from <nplcit id="ncit0001" npl-type="s"><text>L. J. Griffiths, C.W. Jim: "An Alternative Approach to Linearly Constrained Adaptive Beamforming", IEEE Transactions on Antennas and Propagation, Vol. AP-30, No. 1, pages 24-34, January 1982</text></nplcit>.

0005In order to cancel echoes and remove reverberation in recorded sound signals, adaptive filters are known in the art. If music is played back in the passenger compartment while the voice of the passenger is recorded, the components in the recorded signal resulting from a music playback may be attenuated using such an adaptive filter. Adaptive filters and algorithms for filter adaption are described in <nplcit id="ncit0002" npl-type="b"><text>E. Häusler, G. Schmidt: "Acoustic Echo and Noise Control - A Practical Approach", John Wiley &amp; Sons, Hoboken, New Jersey, USA, 2004</text></nplcit>. The components from music playback are removed from the recorded signal by generating a compensation signal, which requires a signal from the music source as a reference signal. The compensation signal is then subtracted from the recorded signal. Such a system is described in the <patcit id="pcit0001" dnum="US6895095B"><text>US patent 6,895,095</text></patcit>. The compensation of interfering components in the recorded sound signal is performed in the frequency domain, which requires extensive signal processing leading to delay. Delay is a major disadvantage in vehicle intercom systems or karaoke systems, since direct sound signals from the speaking/ singing passenger and the sound from the intercom/karaoke system arrive at different times at the listening passengers leading to incomprehensibility. Furthermore, the system does not compensate interfering signals for which no source signal is available. Furthermore, filter adaption in those systems is a problem for highly correlated sound signals such as voice signals. Present filtering techniques can therefore not provide a high quality sound signal that has been compensated for audio components and reverb/echo that has little time delay and can be used in vehicle intercom and karaoke systems.

0006<patcit id="pcit0002" dnum="US6748086B1"><text>US 6,748,086 B1</text></patcit> discloses a cabin communication system for improving clarity of a voice spoken within an interior cabin comprising a microphone array and a processor for combining audio signals to define a beam pattern of the microphone array that includes a plurality of lobes and a plurality of nulls.

0007<patcit id="pcit0003" dnum="US6674865B1"><text>US. 6,674,865 B1</text></patcit> discloses an automatic gain control for a cabin communication system for improving clarity of a voice spoken within a movable interior cabin. The system comprises a filter for removing ambient noise from a recorded audio signal and an acoustic echo canceller for providing an echo-canceled audio signal by making use of a supplied dither signal.

0008Accordingly, there is a need to improve the sound output of vehicle multimedia systems, in particular to remove reverberation and feedback from detected sound signals while ensuring a good audibility of the sound output even at different background noise levels.

Summary of the Invention

0009This need is met by the features of the independent claims. In the dependent claims, preferred embodiments of the invention are described.

0010According to the invention, a method is provided for processing sound signals in a vehicle multimedia system, the method comprising the steps of detecting, by a microphone, a sound signal in a vehicle, the sound signal comprising a first voice signal component corresponding to a voice signal from a passenger of the vehicle, and a second voice signal component resulting from reproducing a processed voice signal by means of a loudspeaker, the sound signal further comprising an audio signal component resulting from reproducing an audio signal of an audio source in the vehicle, estimating an impulse response between a loudspeaker used to reproduce the audio signal and the microphone, simulating a simulated audio signal component based on the impulse response, simulating a simulated second voice signal component based on the impulse response, removing the simulated audio signal component and the simulated second voice signal component from the detected sound signal for obtaining an audio- and echo-compensated signal, estimating a background noise level based on the audio- and echo-compensated signal, and adjusting a volume of sound put out by the vehicle multimedia system based on the estimated background noise level. The audio source may for example be a radio, a CD player or a navigation system, which produce audio signals that are reproduced by a loudspeaker in the vehicle passenger compartment. The audio signal may thus be composed of music or the voice of an announcer. If a passenger in the passenger compartment speaks or sings, his voice is recorded by a microphone as a first voice signal component. This signal component may then for example be given out in a vehicle intercom system to the other passengers and accordingly these components are then recorded again by the microphone as second voice signal components. It is an advantage of the invention that the audio signal component and the second voice signal component are strongly attenuated in the detected sound signal by removing the simulated audio signal component and the simulated second voice signal component from the detected sound signal. Thus, feedback and echo/reverberation are reduced or avoided. On the basis of the audio- and echo-compensated signal the background noise level can now be estimated. In general, the microphone in the passenger compartment records background noise. For example in the case when music is played back by the vehicle multimedia system and no passenger is speaking, the audio- and echo-compensated signal will mainly contain components from background noise. It is thus an advantage to estimate a background noise level based on the audio- and echo-compensated signal. The volume of sound put out by the vehicle multimedia system is then adjusted based on the estimated background noise level. Preferentially, the volume of the audio- and echo-compensated signal is adjusted, e.g. by amplification, wherein the processed voice signal is obtained, which is then reproduced by the loudspeaker. Yet amplification and thus volume of the audio signal may also be adjusted, or the volume of a combination of both signals may be adjusted. This has the advantage that well audible sound is provided to the passengers of a vehicle, even if the background noise increases during the course of their journey. If the background noise level increases, the volume of the sound output of the multimedia system is also increased, so that the sound is well audible above the background noise. If the background noise level decreases again, for example when the car stops at a traffic light, the volume of sound output is also decreased. This process may occur individually for each passenger, and the volume of sound output may additionally be controlled by the vehicle passengers. With the method according to the invention, the vehicle passengers are provided with an echo/reverberation and audio compensated, well audible sound. The audio and echo compensation is relatively simple and does not require extensive processing, whereby little delay is introduced, which is an advantage for vehicle intercom systems and karaoke systems. The method according to the invention can also compensate reverberation of highly correlated signals, such as recorded voice signals.

0011According to an aspect of the invention the impulse response is estimated by an adaption process based on the audio signal and the detected sound signal using a gradient method. The gradient method will be described in detail later. Since the audio signal is generally not highly correlated and is available as a reference signal, it is advantageous to use it to estimate the impulse response. Since the impulse response inside the passenger compartment between the loudspeaker and the microphone may change, for example due to a movement of the passengers, it is advantageous to estimate it by an adaption process. Preferably, the simulated audio signal component is simulated based on the impulse response and the audio signal, wherein a first error signal is obtained by subtracting the simulated audio signal component from the detected sound signal. The first error signal obtained in that way can then be used for the adaption process for estimating the impulse response. In particular, the simulated audio signal component is obtained by a convolution of the impulse response with the audio signal. Preferably, the simulated second voice signal component is simulated based on the impulse response and the processed voice signal, wherein the audio- and echo-compensated signal is obtained as a second error signal by subtracting the simulated second voice signal component from the first error signal. The processed voice signal may be given out by the loudspeakers of the multimedia system, it is thus recorded by the microphone in form of the second voice signal component. It is thus an advantage to base the simulation of the simulated second voice signal component on the processed voice signal. Particularly, the simulated second voice signal component may be obtained from a convolution of the processed voice signal and the impulse response. Using the previously determined impulse response, which was based on the audio signal, has the advantage that it is not necessary to estimate an impulse response based on a highly correlated voice signal from a passenger. Particularly, if the processed voice signal and the audio signal are reproduced by the same loudspeaker, the impulse response for the audio signal and the processed voice signal will be essentially similar, thus leading to good results for the simulated second voice signal component. After subtracting the simulated second voice signal component from the first error signal, a second error signal is obtained in which the second voice signal component and the audio signal component are essentially compensated.

0012In another embodiment of the invention, the method further comprises the steps of determining a short-time power of the detected sound signal, the first error signal and the second error signal, weighting the impulse response with the first weighting factor smaller than 1 if the first error signal short-time power is larger than the detected sound signal short-time power multiplied by a predetermined factor, and weighting the impulse response with the second weighting factor smaller than 1 and smaller than the first weighting factor if the second error signal short-time power is larger than the detected sound signal short-time power multiplied by a predetermined factor. That way the adaptive filtering by means of the estimated impulse response can be stabilized. Since the impulse response is estimated by an adaption process based on the detected sound signal from which the simulated audio signal component has been removed, feedback is provided for the adaption process. In some instances, the adaption of the impulse response may become unstable. This is detected by comparing the short-time power of the first or the second error signal with the short-time power of the detected sound signal multiplied by short-time power of a factor. The factor may for example be between one decibel and three decibel. If the short-time power of the first error signal is larger than the short-time power of the detected sound signal, it can be concluded that the filter adaption has become unstable, since the short-time power of the first error signal should be smaller than the short-time power of the detected sound signal in normal operation. In that case, the estimated impulse response is multiplied by a factor smaller than 1 in order to prevent inflation of the impulse response. In case even the short-time power of the second error signal is larger than the short-time power of the detected sound signal, it can be concluded that even more damping needs to be provided for the impulse response, and it is multiplied by a second weighting factor which is even smaller than the first weighting factor. It is an advantage of the present embodiment that an inflation of the estimated impulse response is damped within a few cycles.

0013The processed voice signal may be directly obtained from the audio- and echo-compensated signal after amplification. Yet the audio- and echo-compensated signal or the processed voice signal may be further processed. Processing may comprise filtering the signal using band pass filters or the like to change the frequency characteristics of the signal and to provide a pleasant sound output for the passengers. It may also comprise noise reduction, automatic gain control, limiting the signal level, frequency equalizing and the like.

0014According to an advantageous embodiment, the background noise level is estimated based on a previously detected noise level, a short-time power of the detected sound signal and a short-time power of the audio- and echo-compensated signal. Preferably, estimating the background noise level comprises the steps of determining a minimum value by taking the smallest value of the following three values: a previously estimated background noise level, a short-time power of the detected sound signal and a short-time power of the audio- and echo-compensated signal, and multiplying the minimum value by a time-dependent factor greater than 1 but smaller than 2. Advantageously, when a short-time power of the audio signal is larger than a predetermined power level, the time-dependent factor is set to a first value and when the short-time power of the audio signal is smaller than said predetermined power level, the time-dependent factor is set to a second value, wherein the first value is smaller than the second value.

0015These steps ensure that a background noise level is accurately estimated. If music is played back inside the passenger compartment, and the background noise level increases inside the vehicle, then the estimated background noise level will increase every cycle since the previously estimated background noise level is multiplied by a factor larger than 1. The estimated background noise level will be increased until it matches the short-time power of the audio- and echo-compensated signal, which essentially corresponds to the recorded background noise. If the background noise drops again, then the audio- and echo-compensated signal will have a lower power than the previously estimated background signal level and the power of the detected sound signal. Accordingly, the new estimated background noise level will drop to the short-time power of the audio- and echo-compensated signal. By these steps, it is ensured that the background noise level is estimated accurately in real time. Using also the short-time power of the detected sound signal has the advantage that in case filter adaptation becomes unstable, leading to an inflation of the impulse response, the estimated background noise level is not inflated. A higher time-dependent factor and thus a faster adaption is chosen when no audio output is present, since an accurate estimation of the background noise level is possible during the many breaks generally present in voice signals.

0016Preferably, the adjustment of the volume of the sound put out by the vehicle multimedia system comprises the steps of amplifying the audio- and echo-compensated signal by an amplification factor based on the estimated background noise level to obtain the processed voice signal, combining the processed voice signal and the audio signal to a combined signal and reproducing the combined signal by means of a loudspeaker. The processed voice signal and the audio signal or the combined signal may be amplified by a power amplifier before reproduction. If several audio channels are provided, then the audio- and echo-compensated signal may be split before amplification into different channels that are later combined with respective audio channels. Each of the audio- and echo-compensated signal channels may be amplified with an individual amplification factor or with the same amplification factor. Preferably, the amplification factor is set to 0 dB when the vehicle is stopped and the motor is turned off. When the vehicle is driving, the background noise level increases with increasing velocity, since engine noise and noise from wind and tires increases with increasing velocity.

0017The estimated background noise level is constantly compared to predetermined noise levels. If the estimated background noise level exceeds one of these predetermined noise levels, the amplification factor is slowly increased to a new value. Preferably, the amplification factor is between 0 and 15 dB. Alternatively, the audio signal may also be amplified before combination, so that the relative power of the two signals can be determined. That may advantageous in situations where one wishes to reduce the volume of the audio output while maintaining the volume of the processed voice signal, whereby the comprehensibility in for example a vehicle intercom system may be improved.

0018In another embodiment of the invention a plurality of microphones and a beamforming unit is provided. A plurality of sound signals is detected and the sound signals are processed using the beamforming unit in order to provide directional detected sound signals. Such a directional detected sound signal may be obtained for each passenger position, wherein multiple microphones may be provided for each passenger position. This has the advantage that background noise and signal components other than the sound originating from a predetermined direction, preferably the direction of the vehicle passenger, are attenuated. Accordingly, the quality of the sound output of the vehicle multimedia system can be further improved. Methods for obtaining a directional detected sound signal from a plurality of detected sound signals by beamforming algorithms is described in detail in <nplcit id="ncit0003" npl-type="s"><text>L. J. Griffiths, C.W. Jim: "An Alternative Approach to Linearly Constrained Adaptive Beamforming", IEEE Transactions on Antennas and Propagation, Vol. AP-30, No. 1, pages 24-34, January 1982</text></nplcit> (document D1). Generally, the different detected sound signals are combined by digital filters in such a way that sound originating from a preferred direction is not attenuated, whereas sound originating from other directions is strongly attenuated. Furthermore, the beamforming unit may comprise a blocking beam former which attenuates the sound originating from the preferred direction, whereas it does not attenuate sound originating from the other direction. By comparing the signal produced by such a blocking beam former with the direction detected sound signal, it can be determined whether a passenger is speaking and which passenger is speaking. If no passenger is speaking, both should have substantially similar short-time powers. If the short-time power of the directional detected sound signal is larger than the short-time power of the sound from the blocking beam former then the passenger associated with the directional detected sound signal is speaking, whereas if it is smaller then another passenger is speaking. Besides improving the sound quality of the detected signal, the beamforming unit has thus the advantage that it can be determined which passenger is speaking or in the case of a karaoke system singing.

0019In another embodiment, the removing of the simulated audio signal component and the simulated second voice signal component is performed before the processing of the detected sound signal for beamforming, and the processing of the detected sound signal for beamforming comprises the step of filtering the sound signal using adaptive filters. By an adaption of the beam former the directivity of the beam former can be further increased. To achieve that, a structure such as the "Griffiths Jim structure", described in D1, can be used. The beam formers can be adapted so that essentially no sound is recorded from the direction of the loudspeakers. Preferably, the filters are adapted when no signal is detected from the preferred direction.

0020According to a preferred embodiment of the invention, a background noise level is estimated for each passenger position and the volume of the sound put out by the vehicle multimedia system is individually adjusted for each passenger position, based on the respective estimated background noise level. Preferably, at least one loudspeaker and at least one microphone are provided for each of at least two vehicle passenger positions, wherein for each position a background noise level is estimated based on the directional detected sound signal for the respective position, an amplification factor is determined based on the estimated background noise level for the respective position, the audio- and echo-compensated signal is amplified using the amplification factors for the respective position and the combined signal is reproduced using the at least one loudspeaker of the respective position. The background noise level may for example be estimated after beamforming. A background noise level can be estimated for each passenger position. The volume of sound put out by the vehicle multimedia system can then be individually adjusted for each passenger position based on the respective estimated background noise level. This is an advantage if different background noise levels exist for different passenger positions, since a well audible sound output can be provided for each passenger position.

0021Furthermore, a passenger position may be determined from which the first voice signal components originate and the volume of sound output of a loudspeaker may be attenuated for said passenger position. This is particularly useful for vehicle intercom systems, where it is not desired to reproduce the voice of the speaking passenger next to the speaking passenger. The position of the passenger from which the first voice signal components originate may be determined by using the beamforming unit. For a loudspeaker associated with that passenger position the output volume of the processed voice signal or the audio signal or both may be attenuated. This is advantageous, since it is often very distracting if a speaking passenger can hear his own voice with a loud volume.

0022According to another aspect of the invention, the multimedia system comprises a karaoke system, wherein the first voice signal component comprises a sound signal created by a singing passenger, and wherein the audio signal originates from a karaoke source. A karaoke system inside a vehicle provides entertainment for passengers, which is particularly useful for long journeys. The karaoke system may comprise a display for displaying information relating to the text of a song, and an audio playback device such as a CD player to play back a karaoke disk comprising music and song text information, e.g. in the CD-G-format. The music generally does not comprise the voice of a singer and the text information is displayed on the display while the music is played back. The passenger can then read the text from the display and sing along with the music which is put out by the vehicle multimedia system. The singing is recorded and then played back via the loudspeakers of the vehicle multimedia system. For such an application the processing of the detected sound signal for removing the audio signal component and the second voice signal component is particularly useful, since reverberation and feedback would strongly decrease the ability to enjoy such a system. In such a system the music output of the karaoke system corresponds to the audio signal which is used as a reference to adapt the estimated impulse response. The first voice signal components would correspond to the voice of a singing passenger, whereas the second voice signal components would correspond to this voice being reproduced by the loudspeakers in the vehicle.

0023According to a further aspect of the invention, at least one function of the multimedia system or the vehicle can be voice-controlled, wherein the processed voice signal is used to control that function. Voice control in general is known in the art. A command spoken by a passenger has to be recognized by a voice control unit, and the function corresponding to the command has to be executed. It is an advantage to use the processed voice signal which has been audio- and echo-compensated, since recognition of the command is facilitated due to the absence of interfering signal components. In such a system the success rate of recognizing the voice command can be increased.

0024The multimedia system may furthermore comprise a communication system such as a telephone system or an vehicle interior communication system using the processed voice signals as an input. Since the processed voice signal contains essentially only first voice signal components, it is an advantage to use this signal as an input for telephone systems, since the person communicating with the vehicle passenger through said telephone system only receives the first voice signal component. Thereby, a clear voice signal is provided to the receiving person. Also the option may be provided that several vehicle passengers participate in the telephone conversation. To that end the utterance of one vehicle passenger may be given out through the multimedia system to the other vehicle passengers. Even so, with the method of the present embodiment, these second voice signal components, which are again recorded by the microphone, do not lead to feedback or reverberation, since they are compensated.

0025Furthermore, the method according to the invention may comprise the step of transferring data between the multimedia system and a multimedia system of a substantially similar kind, wherein the transferred data comprises at least the processed voice signal and wherein the data transfer is wireless. Similarly, a processed voice signal may be received by the multimedia system, wherein the received processed voice signal was sent by the multimedia system of another vehicle. If the multimedia system is used as an intercom system, conversation between passengers of different vehicles becomes possible. That way passengers of one vehicle can communication with passengers of another vehicle without the need for a telephone system. If the multimedia system is used as a karaoke system, transferring data comprising at least the processed voice signal has the advantage that the passengers of the other vehicle that has a multimedia system capable of receiving the data can listen to the singing of the passengers of the first vehicle. Preferably, the transferred data further comprises data relating to the audio signal or the audio signal itself. That means that in a karaoke system the voice of the singing passenger is transferred together with the music given out by the karaoke system. This has the advantage that the passengers of the other vehicle can listen to the singing and the music without the need to have the same music available in their vehicle. Preferably, a stereo signal is transmitted to ensure good music quality. Alternatively, data relating to the audio signal is transmitted together with the processed voice signal. The data related to the audio signal may comprise the interpreter of the song and the song title, or other information that enables the identification of the particular song. If the song is supposed to be played back in the other vehicle together with the transmitted processed voice signal, the multimedia system of the other vehicle must have that song available. This embodiment has the advantage that only a limited amount of data has to be transferred, while a high sound quality is provided in both vehicles. Such systems may furthermore exchange information relating to timing. Due to the processing and the transmission of the processed voice signal, a delay may be introduced. This delay would have to be compensated by the receiving multimedia system, so that the reproduced audio signal and the reproduced processed voice signal have the right timing. For transmission, a mobile phone codec, wireless local area networks codec or broadband Bluetooth codec may be used. With the latter, the two vehicles sending and receiving the signal may have only a limited distance from each other. Nevertheless, a higher transmission bandwidth is possible in that case.

0026Furthermore, a system for processing sound signals in a vehicle multimedia system is provided, comprising at least one loudspeaker for reproducing an audio signal of an audio source and for reproducing a processed voice signal, at least one microphone for detecting a sound signal comprising a first voice signal component corresponding to a voice signal from a passenger of the vehicle, a second voice signal component corresponding to the reproduced processed voice signal, and an audio signal component corresponding to the reproduced audio signal, means for estimating an impulse response between the loudspeakers used to reproduce the audio signal and the microphone, first processing means formed to compensate the audio signal component in the detected sound signal wherein a first error signal is obtained, comprising a first filter using the impulse response as filter coefficients, second processing means formed to compensate the second voice signal component in the detected sound signal wherein a second error signal is obtained, comprising a second filter using the same impulse response as filter coefficients, processing means for estimating the background noise level based on the second error signal, and controlling means for controlling amplification of the second error signal based on the background noise level. Such a system has the advantage that an audio- and echo-compensated signal with the proper volume is provided to the vehicle passenger. Such a system achieves good echo compensation even for highly correlated voice signals. The system can give out well audible sounds to the passengers of a vehicle even if the background noise in the passenger compartment changes. The audio source may be multichannel, such as a stereo or dolby surround audio source. Multiple loudspeakers and multiple microphones may be provided, and these loudspeakers and microphones may be associated with vehicle passenger positions. The system may furthermore comprise signal processing components for processing the detected sound signal after filtering. Preferably, the system comprises a plurality of microphones and a beamforming unit for processing the sound signals detected by said microphones for generating a directional detected sound signal. If several microphones are provided for each passenger position, several directional detected sound signals may be generated, each associated with a particular passenger position. Directional detected sound signals have the advantage that interfering sound signal components originating from other directions than a preferred direction, preferably the direction of the speaker, are strongly attenuated. The system may furthermore comprise decorrelation filters for filtering the detected sound signal and the audio signal. Using decorrelation filters has the advantage that the convergence of the filter adaption is sped up. Filtering may occur in the frequency domain, but preferably, filtering occurs in the time domain to avoid delay. Alternatively, processing may occur in the subband domain.

0027The system may furthermore comprise a stability control unit weighting the impulse response based on a predetermined criterion. Such a criterion may be whether the short-time power of the first error signal is larger than the short-time power of the detected sound signal or whether the short-time power of the second error signal is larger than the short-time power of the detected sound signal or a combination thereof. The impulse response is preferably weighted with a factor smaller than 1 but larger than 0. If the system runs stably, the short-time power of the first error signal should be smaller than the short-time power of the detected sound signal, since the audio signal component has been compensated in the first error signal. In the case of a larger first error signal short-time power the system runs unstable, and the impulse response is weighted with a factor smaller than 1, whereby a damping is provided for the impulse response adaption, preventing an inflation of the impulse response and restoring the stability of the system.

0028Preferably, the means for controlling the amplification are formed for independently controlling sound volume for different passenger positions inside the vehicle, the amplification being based on the voice activity of a passenger on the respective position and/or the background noise present at the respective position. The output of the processed voice signal and the audio signal may be independently controlled, or it may be controlled together. By this feature, a pleasant sound output is generated for each passenger, since at each passenger position different background noise may prevail. Furthermore, if voice activity of the passenger is detected; the output of the processed voice signal can be attenuated for the passenger position, whereby it is avoided that the passenger has to listen to his voice with a high volume, which would be highly distracting.

0029The system may furthermore comprise a karaoke system. Karaoke systems are well-known in the art. They generally comprise a display for displaying the text of the song, and means for playing back the song. The song and the text may for example be stored on a CD or a DVD. The song played back by the karaoke system generally only comprises the music without the singing of the artist. The music together with the singing of the passenger which was detected by the microphone and processed to remove reverberation and the audio components may then be played back by the vehicle multimedia system. The system may furthermore comprise a voice control unit for voice-controlling at least one function of the multimedia system or the vehicle based on the processed voice signal. Thereby, recognition of the voice command spoken by a passenger is facilitated, since interfering signal components are removed by processing the detected sound signal. The system may further comprise a data transfer unit for transferring data comprising the processed voice signal to a vehicle multimedia system capable of receiving said data. This way, intervehicle communication is enabled and intervehicle karaoke sessions become possible. The multimedia system may furthermore be formed as to perform the steps of one of the above-described methods.

0030The invention is illustrated in further detail below with the aid of exemplary embodiments and by referring to the figures.

Brief Description of the Drawings

0031<ul id="ul0001" list-style="none"><li><figref idref="f0001">Fig. 1</figref> shows a schematic drawing of a system for processing sound signals in a vehicle multimedia system.</li><li><figref idref="f0002">Fig. 2</figref> shows a directional detected sound signal and an estimated background noise level.</li><li><figref idref="f0003">Fig. 3</figref> shows a flow diagram of a method for processing sound signals in a vehicle multimedia system.</li><li><figref idref="f0004">Fig. 4</figref> shows an alternate embodiment of a system for processing sound signals in a vehicle multimedia system, comprising a beamforming unit, a stability control unit, a voice control unit and a data transfer unit.</li><li><figref idref="f0005">Fig. 5</figref> is a schematic drawing of a karaoke system.</li></ul>

Detailed Description of the Preferred Embodiments

0032<figref idref="f0001">Fig. 1</figref> represents a system for processing sound signals in a vehicle multimedia system. A vehicle passenger compartment 1, in which a passenger 2 is located, comprises a loudspeaker 3 and a microphone 4. An utterance by the passenger 2 is recorded by the microphone 4 and reproduced by the loudspeaker 3. Loudspeaker 3 furthermore reproduces an audio signal <i>x_e,i(n)</i> of an audio source. As the audio source file may comprise multiple audio channels, the index i is introduced representing the audio channel (e.g. i equals 0, 1 for a stereo signal). The variable n refers to the time. The reproduced audio and voice signals are also recorded by the microphone 4. Thus, a sound signal <i>u(n)</i> detected by the microphone 4 comprises a first voice signal component from the passenger, a second voice signal component from reproducing the passenger voice by the loudspeaker 3, and an audio signal component from reproducing the audio signal of the audio source 5 by the loudspeaker 3. The detected sound signal <i>u(n)</i> is then processed by first processing means 6 to compensate the audio signal component in <i>u(n)</i>. The signal paths of the audio signal from the loudspeaker 3 to the microphone 1 can be described by an impulse response <i>h_i</i>. Means for estimating that impulse response are provided within the first processing means 6. The first processing means 6 furthermore comprise a first adaptive filter 7, using the audio signal <i>x_e,i(n)</i> as an input. The estimation of the impulse response and the compensation of the audio signal component will be described in detail hereafter.

0033The impulse response is estimated by an adaption process. The first adaptive filter 7 simulates the audio signal component by a convolution of the audio signal <i><b>x</b>_e,i(n)</i> and the estimated impulse response <b><i>ĥ</i></b><i>_i</i>, resulting in a simulated audio signal component <i>d̂_e</i>(<i>n</i>). That component is subtracted by a first subtraction unit 8 from the detected sound signal u(n), wherein a first error signal e(n) is obtained: <maths id="math0001" num="(1)"><math display="block"><mi>e</mi><mfenced><mi>n</mi></mfenced><mo>=</mo><mi>u</mi><mfenced><mi>n</mi></mfenced><mo>-</mo><mstyle displaystyle="true"><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover></mstyle><msub><mover><mi>d</mi><mo>^</mo></mover><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><mi>u</mi><mfenced><mi>n</mi></mfenced><mo>-</mo><mstyle displaystyle="true"><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover></mstyle><msubsup><mover><mi>h</mi><mo>^</mo></mover><mi>i</mi><mi>T</mi></msubsup><mfenced><mi>n</mi></mfenced><mo>⁢</mo><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0001.tif" /></maths>

0034The summation here is performed over the audio channels, the parameter M denotes the number of channels. The vector <i><b>x</b>_e,i(n)</i> denotes the vector of the last N audio signals: <maths id="math0002" num="(2)"><math display="block"><msub><mi mathvariant="bold-italic">x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><msup><mfenced open="[" close="]"><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>,</mo><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>,</mo><mo>…</mo><mo>,</mo><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mi>N</mi><mo>+</mo><mn>1</mn></mfenced></mfenced><mi>T</mi></msup></math><img file="EP2018034B1_D0002.tif" /></maths>

0035The estimated impulse responses <b><i>ĥ</i></b><i>_i(n)</i> are subsequently adapted in the direction of a negative gradient of a suitably chosen cost function <i>J(e(n)):</i><maths id="math0003" num="(3)"><math display="block"><msub><mover><mi mathvariant="bold-italic">h</mi><mo>^</mo></mover><mi>i</mi></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>+</mo><mn>1</mn></mfenced><mo>=</mo><msub><mover><mi mathvariant="bold-italic">h</mi><mo>^</mo></mover><mi>i</mi></msub><mfenced><mi>n</mi></mfenced><mo>-</mo><msub><mi>μ</mi><mi>i</mi></msub><mfenced><mi>n</mi></mfenced><mo>⁢</mo><msub><mo>▿</mo><mi>h</mi></msub><mo>⁢</mo><mi>J</mi><mfenced><mi>e</mi><mfenced><mi>n</mi></mfenced></mfenced></math><img file="EP2018034B1_D0003.tif" /></maths>

0036The step size factor µ<i>_i(n)</i> is an important parameter for controlling the adaption. If the vehicle multimedia system is for example used as a karaoke system, it is generally known at which time a passenger will start to sing. This knowledge may be used to decrease the step size factor before the singing starts in order to enable a better adaption. Further information about controlling the step size can be found in <nplcit id="ncit0004" npl-type="b"><text>E. Häusler, G. Schmidt: "Acoustic Echo and Noise Control - A Practical Approach", John Wiley &amp; Sons, Hoboken, New Jersey, USA, 2004</text></nplcit>. For adaption of the estimated impulse responses minimizing the mean square error <maths id="math0004" num="(4)"><math display="block"><mi>J</mi><mfenced><mi>e</mi><mfenced><mi>n</mi></mfenced></mfenced><mo>=</mo><mi>E</mi><mfenced open="{" close="}"><msup><mi>e</mi><mn>2</mn></msup><mfenced><mi>n</mi></mfenced></mfenced></math><img file="EP2018034B1_D0004.tif" /></maths> is a suitable method.

0037In this equation E denotes the expectation value. In this case the adaption of the impulse responses can occur as follows: <maths id="math0005" num="(5)"><math display="block"><msub><mover><mi mathvariant="bold-italic">h</mi><mo>^</mo></mover><mi>i</mi></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>+</mo><mn>1</mn></mfenced><mo>=</mo><msub><mover><mi mathvariant="bold-italic">h</mi><mo>^</mo></mover><mi>i</mi></msub><mfenced><mi>n</mi></mfenced><mo>+</mo><msub><mi>μ</mi><mi>i</mi></msub><mfenced><mi>n</mi></mfenced><mo>⁢</mo><mfrac><mrow><msub><mi>e</mi><mi>i</mi></msub><mfenced><mi>n</mi></mfenced><mo>⁢</mo><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced></mrow><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover><msubsup><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow><mi>T</mi></msubsup><mfenced><mi>n</mi></mfenced><mo>⁢</mo><msub><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced></mrow></mfrac></math><img file="EP2018034B1_D0005.tif" /></maths>

0038It should be noted that in the case of multiple channels an extended normalization has to be performed. Details of such a normalization can be found in <nplcit id="ncit0005" npl-type="b"><text>S. Haykin: "Adaptive Filter Theory", Vol. 4, Prentice Hall, Englewood Cliffs, New Jersey, 2002</text></nplcit>. Additionally, methods to increase the speed of conversion may be used, for example fixed or adaptive decorrelation filters. The adaption of the impulse response may also be implemented in the frequency or subband domain, yet the convolution should occur in the time domain to avoid delay.

0039After compensation of the audio signal component the second voice signal component that is still present in the first error signal is compensated by a second processing means 9. The second processing means 9 comprise a second filter 10 using the estimated impulse response <b><i>ĥ</i></b><i>_i(n)</i> as filter coefficients. The second filter 10 uses the processed voice signal <i>x_r,i(n)</i> as an input. The vector <maths id="math0006" num="(6)"><math display="block"><msub><mi>x</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><msup><mfenced open="[" close="]"><msub><mi>x</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>,</mo><msub><mi>x</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>,</mo><mo>…</mo><mo>,</mo><msub><mi>x</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mi>N</mi><mo>+</mo><mn>1</mn></mfenced></mfenced><mi>T</mi></msup></math><img file="EP2018034B1_D0006.tif" /></maths> contains the last N processed voice signals. The simulated second voice signal component is then obtained by multiplying the impulse response vector with the vector of the last processed voice signals, which corresponds to a discrete convolution: <maths id="math0007" num="(7)"><math display="block"><msub><mover><mi>d</mi><mo>^</mo></mover><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><msubsup><mover><mi>h</mi><mo>^</mo></mover><mi>i</mi><mi>T</mi></msubsup><mfenced><mi>n</mi></mfenced><mo>⁢</mo><msub><mi>x</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0007.tif" /></maths>

0040As mentioned before, an impulse response is calculated for each of the M audio channels, and a simulated second voice signal component is obtained accordingly for each channel. The sum of the simulated second voice signal components can then be subtracted from the first error signal <i>e(n)</i> by a second subtracting unit 11 to obtain the second error signal <i>ẽ</i>(<i>n</i>). Note that only one audio channel is shown in <figref idref="f0001">Fig. 1</figref>. The transfer of the estimated impulse response from the first filter 7 to the second filter 10 is illustrated by an arrow. Instead of transferring the estimated impulse response, both filters may simply access the same memory in which the impulse response is stored. In summary, the audio- and echo-compensated sound signal in form of the second error signal is obtained from the detected microphone signal by the following subtraction: <maths id="math0008" num="(8)"><math display="block"><mover><mi>e</mi><mo>˜</mo></mover><mfenced><mi>n</mi></mfenced><mo>=</mo><mi>u</mi><mfenced><mi>n</mi></mfenced><mo>-</mo><mstyle displaystyle="true"><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover></mstyle><msub><mi>d</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mo>-</mo><mstyle displaystyle="true"><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover></mstyle><msub><mi>d</mi><mrow><mi>r</mi><mo>,</mo><mi>i</mi></mrow></msub><mfenced><mi>n</mi></mfenced><mn>.</mn></math><img file="EP2018034B1_D0008.tif" /></maths>

0041The second error signal <i>ẽ</i>(<i>n</i>) is then processed by signal processing components 12. The components 12 are optional and may comprise noise reduction, equalizers, limiters, automatic gain control and the like. The processed voice signal <i>x_r,i(n)</i> is obtained by splitting the second error signal in M channels corresponding to the M audio channels and amplifying said channels with the same or individual amplification factors. Splitting may occur in the processing components or in the controlling means 14. The processed voice signal is then reproduced by loudspeaker 3. One has to note that generally audio signals are reproduced by the front as well as the back speakers in a multichannel multimedia system, whereas the processed voice signal is only reproduced by the loudspeakers close to the listening passengers. This is not accounted for in equation 8. This is still considered by the method, as the output of the processed voice signal is a priori set to 0 for the channels corresponding to the speaker.

0042The vehicle multimedia system 100 furthermore comprises processing means 13 for estimating a background noise level. The processing means 13 therefore comprise means to estimate a short-time power of the detected sound signal <i>p_u(n)</i> and for determining a short-time power of the audio- and echo-compensated signal ρ<i>_e</i>(<i>n</i>) (or second error signal). The short-time power may be determined by recursive filtering of the squares of the respective signal: <maths id="math0009" num="(9)"><math display="block"><msub><mi>p</mi><mi>u</mi></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><mi mathvariant="italic">λ</mi><mo>⁢</mo><msub><mi>p</mi><mi>u</mi></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>+</mo><mfenced><mn>1</mn><mo>-</mo><mi mathvariant="italic">λ</mi></mfenced><mo>⁢</mo><msup><mi>u</mi><mn>2</mn></msup><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0009.tif" /></maths><maths id="math0010" num="(10)"><math display="block"><msub><mi>p</mi><mover><mi>e</mi><mo>˜</mo></mover></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><msub><mi mathvariant="italic">λp</mi><mover><mi>e</mi><mo>˜</mo></mover></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>+</mo><mfenced><mn>1</mn><mo>-</mo><mi mathvariant="italic">λ</mi></mfenced><mo>⁢</mo><msup><mover><mi>e</mi><mo>˜</mo></mover><mn>2</mn></msup><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0010.tif" /></maths>

0043The time constant λ of the IIR filtering may be chosen from the range of values λ ∈ [0.95, 0.999].

0044The background noise level <i>p_b(n)</i> is then estimated as follows: <maths id="math0011" num="(11)"><math display="block"><msub><mi>p</mi><mi>b</mi></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><mi>min</mi><mfenced open="{" close="}"><msub><mi>p</mi><mi>b</mi></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>,</mo><msub><mi>p</mi><mi>u</mi></msub><mfenced><mi>n</mi></mfenced><mo>,</mo><msub><mi>p</mi><mover><mi>e</mi><mo>˜</mo></mover></msub><mfenced><mi>n</mi></mfenced></mfenced><mo>⁢</mo><mfenced><mn>1</mn><mo>+</mo><mi>ε</mi><mfenced><mi>n</mi></mfenced></mfenced></math><img file="EP2018034B1_D0011.tif" /></maths>

0045The time-dependent factor ε<i>(n)</i> takes on the following values: <maths id="math0012" num="(12)"><math display="block"><mi>ε</mi><mfenced><mi>n</mi></mfenced><mo>=</mo><mfenced open="{" close="}"><mtable><mtr><mtd><msub><mi>ε</mi><mn>0</mn></msub><mo>,</mo><mi mathvariant="italic">if</mi><mspace width="1em" /><msub><mi>p</mi><msub><mi>x</mi><mi>ε</mi></msub></msub><mfenced><mi>n</mi></mfenced><mo>></mo><msub><mi>p</mi><mn>0</mn></msub></mtd></mtr><mtr><mtd><msub><mi>ε</mi><mn>1</mn></msub><mo>,</mo><mi mathvariant="italic">otherwise</mi></mtd></mtr></mtable></mfenced></math><img file="EP2018034B1_D0012.tif" /></maths>

0046The constants ε<i>₀</i> and ε<i>₁</i> are chosen slightly larger than 0, wherein <maths id="math0013" num="(13)"><math display="block"><msub><mi>ε</mi><mn>0</mn></msub><mo>&lt;</mo><msub><mi>ε</mi><mn>1</mn></msub><mn>.</mn></math><img file="EP2018034B1_D0013.tif" /></maths>

0047<i>p_{x<sub2>e</sub2>}</i>(<i>n</i>) is the short-time power of the audio signals, averaged over time and the number of channels M. It can be calculated as follows: <maths id="math0014" num="(14)"><math display="block"><msub><mi>p</mi><msub><mi>x</mi><mi>e</mi></msub></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><mi mathvariant="italic">λ</mi><mo>⁢</mo><msub><mi>p</mi><msub><mi>x</mi><mi>e</mi></msub></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>+</mo><mfenced><mn>1</mn><mo>-</mo><mi mathvariant="italic">λ</mi></mfenced><mo>⁢</mo><mfrac><mn>1</mn><mi>M</mi></mfrac><mstyle displaystyle="true"><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mi>M</mi><mo>-</mo><mn>1</mn></mrow></munderover></mstyle><msubsup><mi>x</mi><mrow><mi>e</mi><mo>,</mo><mi>i</mi></mrow><mn>2</mn></msubsup><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0014.tif" /></maths>

0048The constant po is chosen so that by comparing it with the average music short-time power it can be determined whether music playback is active. Accordingly, if music playback is active, the estimation of the background noise level is more slowly adapted. An estimated background noise level together with a detected sound signal <i>u(n)</i> is shown in <figref idref="f0002">Fig. 2</figref>. The detected sound signal 201 shows a plurality of peaks during times when both music and voice are active, whereas it shows fewer peaks when only the voice is active, and almost no peaks are visible without any music or voice activity.

0049The estimated background noise level is shown as reference sign 202 (bold line). Naturally, the background noise estimation is most accurate when no audio or voice components are present, yet is still achieves good accuracy for the other situations. Although only one microphone and audio channel is shown in <figref idref="f0001">Fig. 1</figref>, microphones may be provided for each passenger position, and accordingly, background noise levels may be estimated for each passenger position. The estimated background noise level is then supplied to controlling means 14 for controlling the amplification of sound output particularly of the processed voice signal in the present embodiments. The control means 14 are furthermore supplied with the second error signal <i>ẽ</i>(<i>n</i>). The control means 14 amplifies the second error signal based on the background noise level. For example, if the vehicle is standing and the engine not running, the user may preset a certain amplification, corresponding to a particular volume of reproduced sound. In that situation the background noise level is close to 0, and no further amplification is provided by the controlling means 14. If the vehicle is now driven at a certain velocity, the engine and the wind and traction noises will cause an increased background noise level. If the background noise level passes certain preset levels, the sound output may be amplified by amplification factors <i>c_i(n)</i>, which may be different for different audio channels corresponding to different passenger positions. Preferably, the amplification is controlled within a range between 0 and 15 dB. The processed voice signal is then combined with the audio signal and put out through the loudspeaker 3. Attention has to be paid that the closed-loop gain does not exceed 1. The amplification may be limited in that case.

0050It should be clear that <figref idref="f0001">Fig. 1</figref> only shows components that are necessary to illustrate an embodiment of the system according to the invention. A multimedia system may comprise many further components which are well-known to a person skilled in the art, such as control elements, display elements, means to receive and demodulate radio signals, CD, DVD, MP3 or other players, communication systems, navigation systems and the like.

0051<figref idref="f0003">Fig. 3</figref> shows a flow diagram of an embodiment of the method according to the invention. In a first step 301, a sound signal is detected by a microphone. The detected sound signal comprises a voice signal component corresponding to a voice signal from a passenger, a second voice signal component corresponding to reproducing a processed voice signal by a loudspeaker and an audio signal component corresponding to an audio signal reproduced by a loudspeaker. In a next step 302, an impulse response between the loudspeaker and the microphone is estimated. If a plurality of audio channels is provided and reproduced by a plurality of loudspeakers, an impulse response is estimated between each loudspeaker and the microphone. Using the estimated impulse response, the audio signal component is simulated in a next step 303. Steps 302 and 303 may be performed in one processing unit, since feedback from the simulated audio signal is provided in order to adapt the estimated impulse response. In a next step 304, the second voice signal component is simulated using the same impulse response that was obtained in step 302. Generally, it is difficult to compensate highly correlated signal components such as voice signal components. The here described method has the advantage that by using the same impulse response even the correlated second voice signal components can be rather easily removed. In a next step 305, the simulated components are removed from the detected sound signal. Preferably, the simulated audio signal component is first subtracted from the detected sound signal to obtain a first error signal, whereafter the simulated second voice signal component is subtracted from the first error signal to obtain an audio- and echo-compensated signal or second error signal <i>ẽ</i>(<i>n</i>). Preferably, the processing is cycle-based, yet it may also be block-based, wherein it should be ensured that small block sizes are used in order to prevent delay. The delay introduced by block-based processing should be below 5-10 ms. Otherwise, the delay between the voice signals emitted by a passenger and the reproduced voice signal will be noticeable by the passenger, which may result in a passenger perceiving the reproduced voice signal as a distracting echo. Particularly in intercom systems and karaoke systems this would be a drawback. Therefore, the signal processing is optimized with respect to delay and processing is mainly performed in the time domain. In a next step 306, the background noise level is estimated based on the audio- and echo-compensated signal. The estimated background noise level is then used in the next step 307 to adjust a volume of sound output by the multimedia system. If the background noise level increases, the volume of sound output is also increased so that the passenger is always provided with a well audible sound.

0052<figref idref="f0004">Fig. 4</figref> shows another embodiment of the vehicle multimedia system according to the invention. Multiple passengers 401 are located in a vehicle passenger compartment 402 which comprises multiple loudspeakers 403 and microphones 404. Multiple microphones are provided for each passenger position, so that beam-formed signals can be obtained for each passenger position. Beamforming takes place in a beamforming unit 405, which is supplied with the multiple detected sound signals. The detected sound signals again comprise first voice signal components, second voice signal components and audio signal components. A directional detected sound signal is obtained for each passenger position by the beamforming unit 405, which may for example comprise a conventional delay and sum beamformer. Alternatively, an adaptive beam former may be used, for example-employing a Griffiths-Jim-structure, yet it is preferably to employ such an adaptive beam former after audio and echo compensation. Several methods of beamforming are known in the art, and several are described in <nplcit id="ncit0006" npl-type="s"><text>L. J. Griffiths, C.W. Jim: "An Alternative Approach to Linearly Constrained Adaptive Beamforming", IEEE Transactions on Antennas and Propagation, Vol. AP-30, No. 1, pages 24-34, January 1982</text></nplcit>. By using a beam-formed signal, audio and echo compensation can be further improved. Furthermore, information about an actively speaking passenger can be derived. The directional detected sound signals are then supplied to a compensation unit 406. In the compensation unit 406, impulse responses between the microphone arrays 404 for each passenger position and the loudspeakers 403 associated with different passenger positions and audio channels are estimated. In particular, an impulse response is estimated for each combination of a directional detected sound signal and each audio channel. For that purpose, a plurality of first processing means as described with reference to <figref idref="f0001">Fig. 1</figref> are provided in the compensation unit 406, wherein said first processing means also compensate the audio signal components in the directional detected sound signals.

0053Furthermore, a plurality of second processing means is provided in the compensation unit 406 for compensating the second voice signal component. The impulse response estimated in the first processing means is adapted by using feedback, whereby the adaption may become unstable. In order to avoid that, a stability control unit 407 is provided. The stability control unit 407 uses the directional detected sound signal, the first error signal and the second error signal as an input. Such a stability control unit may of course be provided for each of the first processing means. In a stability control unit, the short-time power of the three input signals is calculated. That can be achieved as previously described using recursive filtering of the squares of the signals: <maths id="math0015" num="(15)"><math display="block"><msub><mi>p</mi><mi>e</mi></msub><mfenced><mi>n</mi></mfenced><mo>=</mo><mi mathvariant="italic">λ</mi><mo>⁢</mo><msub><mi>p</mi><mi>e</mi></msub><mo>⁢</mo><mfenced><mi>n</mi><mo>-</mo><mn>1</mn></mfenced><mo>+</mo><mfenced><mn>1</mn><mo>-</mo><mi mathvariant="italic">λ</mi></mfenced><mo>⁢</mo><msup><mi>e</mi><mn>2</mn></msup><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0015.tif" /></maths>

0054The time constant λ of the IIR-filtering is preferably chosen from the range λ ∈ [0.95, 0.99]. Instability is detected in two stages. In a first step, the short-time power of the directional detected sound signal <i>p_u(n)</i> is compared to the short-time power of the first error signal <i>p_e(n)</i>. If <maths id="math0016" num="(16)"><math display="block"><msub><mi>p</mi><mi>e</mi></msub><mfenced><mi>n</mi></mfenced><mo>></mo><msub><mi mathvariant="italic">Kp</mi><mi>u</mi></msub><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0016.tif" /></maths> wherein K is a constant between 1 and 3 dB, the estimated impulse response is weighted with a first factor 0 < ρ<i>_e</i> <1. If the short-time power of the second error signal <i>p_ẽ</i>(<i>n</i>) is also larger than the short-time power of the directional detected sound signal <maths id="math0017" num="(17)"><math display="block"><msub><mi>p</mi><mover><mi>e</mi><mo>˜</mo></mover></msub><mfenced><mi>n</mi></mfenced><mo>></mo><msub><mi mathvariant="italic">Kp</mi><mi>u</mi></msub><mfenced><mi>n</mi></mfenced></math><img file="EP2018034B1_D0017.tif" /></maths> the impulse response is weighted with a second factor ρ<i>_ẽ</i> : <maths id="math0018" num="(18)"><math display="block"><mn>0</mn><mo><</mo><msub><mi>ρ</mi><mover><mi>e</mi><mo>˜</mo></mover></msub><mo><</mo><msub><mi>ρ</mi><mi>e</mi></msub><mo><</mo><mn>1</mn></math><img file="EP2018034B1_D0018.tif" /></maths>

0055The damping of the estimated impulse response is stronger in that case, since it is multiplied by two factors smaller than 1. By these measures, an inflation of the estimated impulse response <i>ĥ_i</i>(<i>n</i>) is prevented, sufficient damping occurs in only a few cycles.

0056Similar to <figref idref="f0001">Fig. 1</figref>, processing means 408 for estimating a background noise level are provided in the multimedia system 400. These processing means 408 are provided for each passenger position, i.e. each directional detected sound signal. The processing means 408 are again provided with the directional detected sound signals and the second error signals. The estimated background noise levels are supplied to controlling means 409. The compensation unit 406 may furthermore comprise signal processing components as described with respect to <figref idref="f0001">Fig. 1</figref>, and it submits an audio- and echo-compensated signal to the controlling means 409. The compensation unit 406 may combine the audio- and echo-compensated signal of the different passenger positions before submission or it may submit an audio- and echo-compensated signal for each passenger position. The controlling means 409 furthermore receive information from the beamforming unit 405, said information relating to speech activity of the passengers. Based on the speech activity of the passengers and the estimated background noise levels the controlling means 409 can control the amplification of the audio- and echo-compensated signals. For each passenger position, a separate amplification factor may be determined based on the estimated background noise level for the respective passenger position. The sound output through the speaker for the respective position is then amplified using the amplification factor. The amplification of the audio signals of different audio channels may be performed independently from the amplification of the processed voice signals based on the estimated background noise levels (not shown). Based on the speech activity information supplied by the beamforming unit 405, the output of the processed voice signal for the speaking passenger may be attenuated. Alternatively, the voice output by the front speakers may be attenuated when a person in the front of the vehicle is speaking, while the voice output of the rear speakers may be attenuated when a person in the rear of the vehicle is speaking. The audio source 410 supplying the audio signal may comprise a karaoke system. In that case music is played back to the passengers and the singing of the passengers is recorded. After audio and echo compensation, the singing is again played back to the passenger through the loudspeakers 403. In such an application the compensation is particularly advantageous, since the processed voice signals may be reproduced by the loudspeakers 403 without attenuation. Providing amplification of the audio- and echo-compensated signal based on the estimated background noise level is also advantageous, since it has been shown that if the background noise level increases by 1 dB, a person tends to sing only 0.3 dB louder, so additional amplification is required.

0057The processed voice signals and the audio signals of the different audio channels are supplied to a power amplifier 414, in which these signals are channelwise combined and power-amplified so that they can be reproduced by the loudspeakers 403. Alternatively, the processed voice signals and the audio signals may be separately amplified and reproduced by separate loudspeakers.

0058Furthermore, the vehicle multimedia system of <figref idref="f0004">Fig. 4</figref> comprises a voice control unit 411. The voice control unit 411 uses the processed voice signal as an input. That way the identification of a voice command can be enhanced. Recognition of the voice command would be much more difficult if interfering audio signal components or an echo from the voice command would be comprised in the input signal to the voice control unit. After recognition of the voice command, the voice control unit 411 may perform a function associated with said voice command. The voice control unit 411 may also be directly supplied with the second error signal.

0059Furthermore, a data transfer unit 412 is provided. The data transfer unit 412 may communicate with another data transfer unit of a similar vehicle multimedia system. The data transfer unit 412 uses the processed voice signal as an input and may transmit data comprising the processed voice signal to a receiving system. The transmission is wireless, using codecs such as a mobile phone codec, wireless local area network codec or broadband Bluetooth codec. Accordingly, the passengers of the vehicle which comprises the receiving multimedia system can listen to the utterances of the passengers 401. Similarly, the data transfer unit 412 may receive data comprising voice signals from the other vehicle multimedia system. This data is then submitted to the compensation unit 406, where it can be processed, used for compensation and given out to the controlling means 409, where it can be amplified and reproduced using the loudspeakers 403. With this arrangement, the received voice signal is also provided to the compensation unit 406 as a reference signal, since these voice components again have to be compensated from the detected sound signals. By means of the data transfer unit 412 passengers of the two vehicles which comprise the sending and receiving multimedia systems can communicate with each other. If the vehicle multimedia system is used as a karaoke system, the singing of the passengers may be transmitted from one vehicle to the other one. That way passengers of both vehicles can sing together or the passengers of one vehicle may listen to the singing of the passengers of the other vehicle. For this purpose information relating to the music given out by the karaoke system has to be transmitted between the vehicles. In a situation where a particular song to which the passengers sing is provided in only one vehicle multimedia system, the audio signal from the karaoke system may be provided to the data transfer unit and may be transferred wirelessly to the multimedia system of the other vehicle. If said song is present in both vehicle multimedia systems, only information relating to the song title and the name of the interpreter has to be exchanged together with information relating to timing. Particularly if passengers of both vehicles want to sing together, both songs have to start playing simultaneously. For this application, it is particularly important to minimize delay. Reference numeral 413 denotes an antenna used to receive and send information by the data transfer unit 412.

0060<figref idref="f0005">Fig. 5</figref> shows a karaoke system 505 which may be part of the vehicle multi-media system 500. The singing of the passenger 501 is recorded by a microphone 502 and played back via loudspeakers 503. The passenger 501 reads the text of the song which he is singing from the display 504. The display may for example be mounted in a headrest or in the dashboard. The karaoke system 505 may comprise a CD/DVD player for playing back karaoke CDs/DVDs. As such, the karaoke system would give out the music to the song the passenger 501 is presently singing through loudspeakers 503 together with the singing of the passenger. The text to the song is displayed by the karaoke system 505 on the display 504. The karaoke system may furthermore comprise all the features known from conventional karaoke systems. Multiple microphones may be provided so that multiple passengers may sing simultaneously, or a music video may be shown together with the text of the song on the display 504.

0061The signal processing means described in this application may be implemented as software on a digital signal processor, they may be provided as separate processing chips, which may for example be implemented on a card that can be connected to the multimedia bus system of a vehicle, or may be provided in other forms known to the person skilled in the art.