ROOM CHARACTERIZATION AND CORRECTION FOR MULTI-CHANNEL AUDIO

• US 20120288124A1
• Filed: 05/09/2011
• Published: 11/15/2012
• Est. Priority Date: 05/09/2011
• Status: Active Grant

• Current Assignee: DTS, Inc. (Adeia Inc.)
• Original Assignee: DTS, Inc. (Adeia Inc.)
• Inventors: Fejzo, Zoran, Johnston, James D.
• Granted Patent: US 9,031,268 B2
• Time in Patent Office: Days
• US Class Current: 381/303
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407
• CPC Class Codes: H04R 5/02 Spatial or constructional arrangements of loudspeakers
• CPC Class Codes: H04S 2400/01 Multi-channel, i.e. more than two input channels, sound reproduction with two speakers wherein the multi-channel information is substantially preserved
• CPC Class Codes: H04S 2420/01 Enhancing the perception of the sound image or of the spatial distribution using head related transfer functions [HRTF's] or equivalents thereof, e.g. interaural time difference [ITD] or interaural level difference [ILD]
• CPC Class Codes: H04S 3/008 in which the audio signals are in digital form, i.e. employing more than two discrete digital channels data reduction aspects thereof based on psychoacoustics G10L19/02
• CPC Class Codes: H04S 7/301 Automatic calibration of stereophonic sound system, e.g. with test microphone
• CPC Class Codes: H04S 7/303 Tracking of listener position or orientation

Abstract

Devices and methods are adapted to characterize a multi-channel loudspeaker configuration, to correct loudspeaker/room delay, gain and frequency response or to configure sub-band domain correction filters.

Claims

1. A method for characterizing a multi-channel loudspeaker configuration, comprising:

producing a first probe signal;

supplying the first probe signal to a plurality of audio outputs coupled to respective electro-acoustic transducers positioned in a multi-channel configuration in a listening environment for converting the first probe signal to a first acoustic response and for sequentially transmitting the acoustic responses in non-overlapping time slots separated by silent periods as sound waves into the listening environment; and

for each said audio output,

receiving sound waves at a multi-microphone array comprising at least two non-coincident acousto-electric transducers, each converting the acoustic responses to first electric response signals;

deconvolving the first electric response signals with the first probe signal to determine a first room response for said electro-acoustic transducer at each said acousto-electric transducer;

computing and recording in memory a delay for said electro-acoustic transducer at each said acousto-electric transducer; and

recording the first room responses in memory for a specified period offset by the delay for said electro-acoustic transducer at each said acousto-electric transducer;

based on the delays to each said acousto-electro transducer, determining a distance and at least a first angle to each said electro-acousto transducer; and

using the distances and at least said first angles to the electro-acousto transducers, automatically selecting a particular multi-channel configuration and computing a position for each electro-acousto transducer in that multi-channel configuration within the listening environment.

2. The method of claim 1, wherein the step of computing the delay comprises:

processing each said first electric response signal and the first probe signal to generate a time sequence;

detecting an existence or absence of a pronounced peak in the time sequence as indicating whether the audio output is coupled to the electro-acoustic transducer; and

computing the position of the peak as the delay.

3. The method of claim 1, wherein the first electric response signal is partitioned into blocks and deconvolved with a partition of the first probe signal as the first electrical response is received at the acousto-electric transducers, and wherein the delay and first room response are computed and recorded to memory in the silent period prior to the transmission of the next probe signal.

4. The method of claim 3, wherein the step of deconvolving the partitioned first response signal with the partition of the first probe signal comprises:

pre-computing and storing a set of K partitioned N-point Fast Fourier Transforms (FFTs) of a time-reversed first probe signal of length K*N/2 for non-negative frequencies as a probe matrix;

computing an N-point FFT of successive overlapping blocks of N/2 samples of the first electrical response signal and storing the N/2+1 FFT coefficients for non-negative frequencies as a partition;

accumulating K FFT partitions as a response matrix;

performing a fast convolution of the response matrix with the probe matrix to provide an N/2+1 point frequency response for the current block;

computing an N-point inverse FFT of the frequency response with conjugate symmetric extension to the negative frequencies to form a first candidate room response for the current block; and

appending the first candidate room responses for successive blocks to form the first room response.

5. The method of claim 4, wherein the step of estimating the delay comprises:

computing an N-point inverse FFT of the frequency response with the negative frequency values set to zero to produce a Hilbert Envelope (HE);

tracking the maximum of the HE over successive blocks to update the computation of the delay.

6. The method of claim 5, further comprising:

supplying a second pre-emphasized probe signal to each of the plurality of audio outputs after the first probe signal to record second electrical response signals;

deconvolving overlapping blocks of the second response signals with the partition of the first probe signal to generate a sequence of second candidate room responses; and

using the delay for the first probe signal to append successive second candidate room responses to form the second room response.

7. The method of claim 1, wherein,

if said multi-microphone array comprises only two acousto-electric transducers, computing at least said first angle to electro-acoustic transducers located on a half-plane;

if said multi-microphone array comprises only three acousto-electric transducers, computing at least said first angle to electro-acoustic transducers located on a plane; and

if said multi-microphone array comprises four or more acousto-electric transducers, computing at least said first angle as an azimuth angle and an elevation angle to electro-acoustic transducers located in three-dimensions.

8. A method for characterizing a listening environment for playback of multi-channel audio, comprising:

producing a first probe signal;

supplying the first probe signal to each of a plurality of electro-acoustic transducers positioned in a multi-channel configuration in a listening environment for converting the first probe signal to a first acoustic response and sequentially transmitting the acoustic responses in non-overlapping time slots as sound waves into the listening environment; and

for each said electro-acoustic transducer,

receiving the sound waves at a multi-microphone array comprising at least two non-coincident acousto-electric transducers each converting the acoustic responses to first electric response signals;

deconvolving the first electric response signals with the first probe signal to determine a room response for each electro-acoustic transducer;

for frequencies above a cut-off frequency, computing a first part of a room energy measure from the room responses as a function of sound pressure;

for frequencies below the cut-off frequency, computing a second part of the room energy measure from the room responses as a function of sound pressure and sound velocity;

blending the first and second parts of the energy measure to provide the room energy measure over the specified acoustic band; and

computing filter coefficients from the room energy measure.

9. The method of claim 8, wherein a processor computes the filter coefficients from the room energy measure.

10. The method of claim 9, further comprising the step of:

using the filter coefficients to configure a digital correction filter in a processor.

11. The method of claim 10, further comprising the steps of:

receiving a multi-channel audio signal;

decoding the multi-channel audio signal with a processor to form an audio signal for each said electro-acoustic transducer;

passing each of said audio signals through the corresponding digital correction filter to form a corrected audio signal; and

supplying each said corrected audio signal to the corresponding electro-acoustic transducer for converting the corrected audio signals to acoustics responses and transmitting the acoustic responses as sound waves into the listening environment.

12. The method of claim 8, further comprising:

progressively smoothing the room responses or the room energy measure so that greater smoothing is applied to higher frequencies.

13. The method of claim 12, wherein the step of progressively smoothing the room responses comprising applying a time varying filter to the room response in which the bandwidth of the low pass response of the filter becomes progressively smaller in time.

14. The method of claim 12, wherein the step of progressively smoothing the room energy measure comprises applying forward and backward frequency domain average with a variable forgetting factor.

15. The method of claim 8, wherein the second part of the energy measure is computed by,

computing a first energy component as a function of sound pressure from the room responses;

computing a pressure gradient from said room responses;

applying a frequency dependent weighting to the pressure gradient to calculate sound velocity components;

computing a second energy component from the sound velocity components; and

computing the second part of the energy measure as a function of the first and second energy components.

16. The method of claim 15, wherein the steps of computing the pressure gradient and applying the frequency dependent weighting to the pressure gradient to calculate the sound velocity components are integrally performed directly from the room responses.

17. The method of claim 15, wherein computing the first energy component comprises:

averaging the room responses for the at least two said acoustic-electric transducers to compute an average frequency response; and

computing the first energy component from the average frequency response.

18. The method of claim 8, wherein said first probe signal is a broadband sequence characterized by a magnitude spectrum this is substantially constant over a specified acoustic band, further comprising:

producing a second probe signal, said second probe signal being a pre-emphasized sequence characterized by a pre-emphasis function with magnitude spectrum inversely proportion to frequency applied to a base-band sequence that provides an amplified magnitude spectrum over a low frequency portion of specified the acoustic band;

supplying the second probe signal to each of the electro-acoustic transducers for converting the second probe signals to second acoustic responses and transmitting the second acoustic responses in non-overlapping time slots as sound waves in the listening environment;

for each said electro-acoustic transducer,

receiving the sound waves at the multi-microphone array for said first and second probe signals with said at least two non-coincident acousto-electric transducers each converting the acoustic responses to first and second electric response signals as a measure of sound pressure;

deconvolving the first and second electric response signals with the first probe signal and the base-band sequence, respectively, to determine first and second room responses for each electro-acoustic transducer;

for frequencies above a cut-off frequency, computing a first part of a room energy measure from the first room responses as a function of sound pressure;

for frequencies below a cut-off frequency, computing a second part of the room energy measure from the second room responses as a function of sound pressure and sound velocity;

blending the first and second parts of the energy measure to provide the room energy measure over the specified acoustic band; and

computing filter coefficients from the room energy measure.

19. The method of claim 18, wherein the broadband sequence is the base-band sequence, said pre-emphasis function being applied to the base-band sequence to generate the pre-emphasized sequence.

20. The method of claim 19, wherein the broadband sequence comprises an all-pass sequence characterized by a magnitude spectrum that is substantially constant over the specified acoustic band and an autocorrelation sequence having a zero-lag value at least 30 dB above any non-zero lag value.

21. The method of claim 20, wherein the all-pass sequence is formed by,

generating a random number sequence between −

π and

+π

;

applying overlapping high and low pass filters to smooth the random number sequence;

generate an all-pass probe signal in the frequency domain having unity magnitude and phase of the smoothed random number sequence;

performing an inverse FFT on the all-pass probe signal to form the all-pass sequence, and wherein the pre-emphasized sequence is formed by

applying the pre-emphasis function to the all-pass probe signal in the frequency domain to form a pre-emphasized probe signal in the frequency domain; and

performing an inverse FFT on the pre-emphasized probe signal to form the pre-emphasized sequence.

22. The method of claim 18, wherein the second part of the energy measure is computed by,

computing a first energy component as a function of sound pressure from the second room responses;

computing a pressure gradient from said second room responses;

computing sound velocity components from the pressure gradient;

computing a second energy component from the velocity components; and

computing the second part of the energy measure as a function of the first and second energy components.

23. The method of claim 22, wherein the first energy component is computed by,

computing an average pre-emphasized frequency response from the second room responses for the at least two said acoustic-electric transducers;

applying a de-emphasis scaling to the pre-emphasized average frequency response; and

computing the first energy component from the average pre-emphasized frequency response.

24. The method of claim 22, wherein the steps of computing the pressure gradient and applying the frequency dependent weighting to the pressure gradient to calculate the sound velocity components are integrally performed directly from the room responses.

25. The method of claim 22, wherein the second part of the energy measure is the sum of the first and second energy components.

26. The method of claim 8, wherein the filter coefficients for each channel are computed by comparing the room energy measure to a channel target curve, further comprising applying frequency smoothing to the room energy measure to define the channel target curve.

27. The method of claim 26, further comprising:

averaging the channel target curves to form a common target curve; and

applying correction to each correction filter to compensate for difference between the channel and common target curves.

28. A method of generating correction filters for a multi-channel audio system, comprising:

providing a P-band oversampled analysis filter bank that downsamples an audio signal to base-band for P sub-bands and a P-band oversampled synthesis filter bank that upsamples the P sub-bands to reconstruct the audio signal where P is an integer;

providing a spectral measure for each channel;

combining each said spectral measure with a channel target curve to provide an aggregate spectral measure per channel;

for at least one channel,

extracting portions of the aggregate spectral measure that correspond to different sub-bands;

remapping the extracted portions of the spectral measure to base-band to mimic the downsampling of the analysis filter bank;

estimating an auto regressive (AR) model to the remapped spectral measure for each sub-band; and

mapping coefficients of each said AR model to coefficients of a minimum-phase all-zero sub-band correction filter; and

configuring P digital all-zero sub-band correction filters from the corresponding coefficients that frequency correct the P base band audio signals between the analysis and synthesis filter banks.

29. The method of claim 28, wherein the spectral measure comprises a room spectral measure.

30. The method of claim 28, wherein the P sub-bands are of uniform bandwidth and overlapping.

31. The method of claim 28, wherein the spectral measure has progressively less resolution at higher frequencies.

32. The method of claim 28, wherein the AR model is computed by,

computing an autocorrelation sequence as an inverse FFT of the remapped spectral measure;

and

applying a Levinson-Durbin algorithm to the autocorrelation sequence to compute the AR model.

33. The method of claim 32, wherein the Levinson-Durbin algorithm produces residual power estimates for the sub-bands, further comprising:

selecting an order for the correction filter based on the residual power estimate for the sub-band.

34. The method of claim 28, wherein the channel target curve is a user selected target curve.

35. The method of claim 28, further comprising applying frequency smoothing to the channel room spectral response to define the channel target curve.

36. The method of claim 28, further comprising:

providing a common target curve for all said channels; and

applying correction to each correction filter to compensate for difference between the channel and common target curves.

37. The method of claim 33, further comprising averaging the channel target curves to form the common target curve.

38. A device for processing multi-channel audio, comprising:

a plurality of audio outputs for driving respective electro-acoustic transducers coupled thereto, said electro-acoustic transducers positioned in a multi-channel configuration in a listening environment;

one or more audio inputs for receiving first electric response signals from a plurality of acousto-electro transducers coupled thereto;

an input receiver coupled to the one or more audio inputs for receiving the plurality of first electric response signals;

device memory, and

one or more processors adapted to implement,

a probe generating and transmission scheduling module adapted to,

produce a first probe signal, and

supply the first probe signal to each of the plurality of audio outputs in non-overlapping time slots separated by silent periods;

a room analysis module adapted to,

for each said audio output, deconvolve the first electric response signals with the first probe signal to determine a first room response at each said acousto-electric transducer, compute and record in the device memory a delay at each said acousto-electric transducer and record the first room responses in the device memory for a specified period offset by the delay at each said acousto-electric transducer,

based on the delays at each said acousto-electro transducer for each said electro-acoustic transducer, determine a distance and at least a first angle to the electro-acousto transducer, and

using distances and at least the first angles to the electro-acousto transducers, automatically select a particular multi-channel configuration and compute a position for each electro-acousto transducer in that multi-channel configuration within the listening environment.

39. The device of claim 38, wherein the room analysis module is adapted to partition the first electric response signal into overlapping blocks and deconvolve each block with a partition of the first probe signal as the first electrical response is received and to compute and record the delay and first room response in the silent period prior to the transmission of the next probe signal.

40. A device for processing multi-channel audio, comprising:

a plurality of audio outputs for driving respective electro-acoustic transducers coupled thereto;

one or more audio inputs for receiving first electric response signals from at least two non-coincident acousto-electro transducers coupled thereto;

an input receiver coupled to the one or more audio inputs for receiving the plurality of first electric response signals;

device memory, and

one or more processors adapted to implement,

a probe generating and transmission scheduling module adapted to,

produce a first probe signal, and

supply the first probe signal to each of the plurality of audio outputs in non-overlapping time slots separated by silent periods;

a room analysis module adapted to, for each said electro-acoustic transducer,

deconvolve the first electric response signals with the first probe signal to determine a room response at each acousto-electric transducer for the electro-acoustic transducer;

for frequencies above a cut-off frequency, compute a first part of a room energy measure from the room responses as a function of sound pressure;

for frequencies below the cut-off frequency, compute a second part of the room energy measure from the room responses as a function of sound pressure and sound velocity;

blend the first and second parts of the energy measure to provide the room energy measure over the specified acoustic band; and

compute filter coefficients from the room energy measure.

41. The device of claim 40, wherein said first probe signal is a broadband sequence characterized by a magnitude spectrum that is substantially constant over a specified acoustic band, and wherein the probe generating and transmission scheduling module is adapted to produce and supply a second probe signal to each of the electro-acoustic transducers, said second probe signal being a pre-emphasized sequence characterized by a pre-emphasis function with magnitude spectrum inversely proportion to frequency applied to a base-band sequence that provides an amplified magnitude spectrum over a low frequency portion of specified the acoustic band, and wherein the analysis module is adapted to convert acoustic responses for the second probe signals into second electric response signals and deconvolve those second electric response signals with the base-band sequence to determine second room responses at each acouto-electric transducer for the electro-acoustic transducer, and for frequencies above the cut-off frequency, compute a first part of the room energy measure from the first room responses as a function of sound pressure and for frequencies below the cut-off frequency, compute the second part of the room energy measure from the second room responses as a function of sound pressure and sound velocity, and blend the first and second parts of the energy measure to provide the room energy measure over the specified acoustic band.

42. The device of claim 41, wherein the analysis module is adapted to compute the second part of the energy measure by,

computing a first energy component as a function of sound pressure from the second room responses;

estimating a pressure gradient from said second room responses;

estimating sound velocity components from the pressure gradient;

computing a second energy component from the sound velocity components; and

computing the second part of the energy measure as a function of the first and second energy components.

43. A device for generating correction filters for a multi-channel audio system,

one or more processors adapted to implement for at least one audio channel,

a playback module adapted to provide a P-band oversampled analysis filter bank that downsamples an audio signal to base-band for P sub-bands, P minimum-phase all-zero sub-band correction filters, and a P-band oversampled synthesis filter bank that upsamples the P sub-bands to reconstruct the audio signal where P is an integer, and

an analysis module adapted to combine a spectral measure with a channel target curve to provide an aggregate spectral measure, extract and remap portions of the aggregate spectral measure that correspond to different sub-bands to base-band to mimic the downsampling of the analysis filter bank, compute an auto regressive (AR) model to the remapped spectral measure for each sub-band, and map coefficients of each said AR model to the coefficients of the corresponding minimum-phase all-zero sub-band correction filter in the playback module.

44. The device of claim 43, wherein the analysis module computes the AR module by,

computing an autocorrelation sequence as an inverse FFT of the remapped spectral measure;

and

applying a Levinson-Durbin algorithm to the autocorrelation sequence to compute the AR model.

45. A method of characterizing a listening environment, comprising:

producing a first probe signal, said first probe signal being a broadband sequence characterized by a magnitude spectrum that is substantially constant over a specified acoustic band and an autocorrelation sequence having a zero-lag value at least 30 dB above any non-zero lag value;

producing a second probe signal, said second probe signal being a pre-emphasized sequence characterized by a pre-emphasis function applied to a baseband sequence that provides an amplified magnitude spectrum over a specified target band that overlaps the specified the acoustic band;

supplying the first and second probe signals to each of a plurality of electro-acoustic converters in a multichannel audio system for converting the first and second probe signals to first and second acoustic responses and sequentially transmitting the acoustic responses in non-overlapping time slots as sound waves in a listening environment; and

for each said electro-acoustic converter,

receiving the sound waves at one or more acousto-electric transducers for converting the acoustic responses to first and second electric response signals;

deconvolving the first and second electric response signals to determine first and a second room responses;

for frequencies outside the target band, computing a first spectral measure from the first room response;

for frequencies in the target band, computing a second spectral measure from the second response;

blending the first and second spectral measures to provide a spectral measure over the specified acoustic band.

46. The method of claim 45, wherein the first probe signal'"'"'s broadband sequence provides the baseband sequence for the second probe signal.