Abstract: An apparatus and method of pre-conditioning audio for machine perception. Machine perception differs from human perception, and different processing parameters are used for machine perception applications (e.g., speech to text processing) as compared to those used for human perception applications (e.g., voice communications). These different parameters may result in pre-conditioned audio that is worsened for human perception yet improved for machine perception.

Abstract: Systems, methods, and computer program products for echo cancellation with prediction filter adaptation and detection of wideband offset between a reference signal (available to the echo canceller) and an output signal (unavailable to the echo canceller), where the output signal has been generated by applying at least one level shift to the reference signal, e.g. such that the level shift is unknown to the echo canceller.

Abstract: An apparatus and method of pre-conditioning audio for machine perception. Machine perception differs from human perception, and different processing parameters are used for machine perception applications (e.g., speech to text processing) as compared to those used for human perception applications (e.g., voice communications). These different parameters may result in pre-conditioned audio that is worsened for human perception yet improved for machine perception.

Abstract: In some embodiments, an echo cancellation method which includes adaptation of at least one prediction filter, with adaptation step size controlled using gradient descent on a set of filter coefficients of the filter, where control of the adaptation step size is based at least in part on a direction of adaptation and a predictability of a gradient of adaptation (e.g., a gradient vector). Other aspects of embodiments of the invention include systems, methods, and computer program products for controlling adaptation step size of adaptive (e.g., low-complexity adaptive) echo cancellation. In some embodiments, adaptation step size control is based on a normalized, scaled gradient of adaptation, or includes smoothing of a normalized gradient of adaptation.

Abstract: Systems, methods, and computer program products for echo cancellation with prediction filter adaptation and detection of wideband offset between a reference signal (available to the echo canceller) and an output signal (unavailable to the echo canceller), where the output signal has been generated by applying at least one level shift to the reference signal, e.g. such that the level shift is unknown to the echo canceller.

Abstract: In some embodiments, an echo cancellation method which includes adaptation of at least one prediction filter, with adaptation step size controlled using gradient descent on a set of filter coefficients of the filter, where control of the adaptation step size is based at least in part on a direction of adaptation and a predictability of a gradient of adaptation (e.g., a gradient vector). Other aspects of embodiments of the invention include systems, methods, and computer program products for controlling adaptation step size of adaptive (e.g., low-complexity adaptive) echo cancellation.

Abstract: An apparatus and method of pre-conditioning audio for machine perception. Machine perception differs from human perception, and different processing parameters are used for machine perception applications (e.g., speech to text processing) as compared to those used for human perception applications (e.g., voice communications). These different parameters may result in pre-conditioned audio that is worsened for human perception yet improved for machine perception.

Abstract: Systems, methods, and computer program products for frequency-domain estimation of latency between audio signals. In some embodiments, the estimation is performed on first blocks of data indicative of samples of a first audio signal and second blocks of data indicative of samples of a second audio signal, and includes determining a coarse latency estimate, including by determining gains which, when applied to some of the second blocks, determine estimates of one of the first blocks, and identifying one of the estimates as having a best spectral match to said one of the first blocks. A refined latency estimate is determined from the coarse estimate and some of the gains. Optionally, at least one metric indicative of confidence in the refined latency estimate is generated. Audio processing (e.g., echo cancellation) may be performed on the frequency-domain data, including by performing time alignment based on the refined latency estimate.

Abstract: An apparatus and method of excursion protection of a loudspeaker. The method includes attenuating selected bands in a transform domain, controlled by a feedback signal resulting from an excursion transfer function that has been modified according to the real-time operational characteristics of the loudspeaker. In this manner, the system reduces the amount of wideband attenuation needed to address the predicted excursion, resulting in a better listening experience.

Abstract: An apparatus and method of excursion protection of a loudspeaker. The method includes attenuating selected bands in a transform domain, controlled by a feedback signal resulting from an excursion transfer function that has been modified according to the real-time operational characteristics of the loudspeaker. In this manner, the system reduces the amount of wideband attenuation needed to address the predicted excursion, resulting in a better listening experience.

Abstract: Systems, methods, and computer program products for frequency-domain estimation of latency between audio signals. In some embodiments, the estimation is performed on first blocks of data indicative of samples of a first audio signal and second blocks of data indicative of samples of a second audio signal, and includes determining a coarse latency estimate, including by determining gains which, when applied to some of the second blocks, determine estimates of one of the first blocks, and identifying one of the estimates as having a best spectral match to said one of the first blocks. A refined latency estimate is determined from the coarse estimate and some of the gains. Optionally, at least one metric indicative of confidence in the refined latency estimate is generated. Audio processing (e.g., echo cancellation) may be performed on the frequency-domain data, including by performing time alignment based on the refined latency estimate.

Abstract: A method of processing audio data for replay on a mobile device with a first speaker and a second speaker, wherein the audio data comprises a respective audio signal for each of the first and second speakers, includes: determining a device orientation of the mobile device; if the determined device orientation is vertical orientation, applying a first processing mode to the audio signals for the first and second speakers; and if the determined device orientation is horizontal orientation, applying a second processing mode to the audio signals for the first and second speakers.

Abstract: In an apparatus configured to perform signal processing on audio data of a first sampling rate, methods disclosed herein comprise receiving audio data of a second sampling rate, the second sampling rate being higher than the first sampling rate. The methods comprise applying filtering to the audio data of the second sampling rate to thereby produce first filtered audio data and second filtered audio data, the first filtered audio data comprising mainly component frequencies which are audible to the human ear, the second filtered audio data comprising mainly components frequencies which are substantially inaudible to the human ear. The methods further comprise applying first signal processing to the first filtered audio data; and applying second signal processing to the second filtered audio data, the second signal processing having a lower computational complexity than the first signal processing. Corresponding apparatus and computer readable media are also disclosed herein.

Abstract: A method of processing audio data for replay on a mobile device with a first speaker and a second speaker, wherein the audio data comprises a respective audio signal for each of the first and second speakers, includes: determining a device orientation of the mobile device; if the determined device orientation is vertical orientation, applying a first processing mode to the audio signals for the first and second speakers; and if the determined device orientation is horizontal orientation, applying a second processing mode to the audio signals for the first and second speakers.

Abstract: In an apparatus configured to perform signal processing on audio data of a first sampling rate, methods disclosed herein comprise receiving audio data of a second sampling rate, the second sampling rate being higher than the first sampling rate. The methods comprise applying filtering to the audio data of the second sampling rate to thereby produce first filtered audio data and second filtered audio data, the first filtered audio data comprising mainly component frequencies which are audible to the human ear, the second filtered audio data comprising mainly components frequencies which are substantially inaudible to the human ear. The methods further comprise applying first signal processing to the first filtered audio data; and applying second signal processing to the second filtered audio data, the second signal processing having a lower computational complexity than the first signal processing. Corresponding apparatus and computer readable media are also disclosed herein.

Abstract: An audio peak limiter apparatus which calculates a smoothed sequence of gains for application to a sequence of blocks of samples of an audio signal. The apparatus sometimes calculates a candidate gain, to replace a too-large smoothed gain, such that applying the candidate gain would produce no scaled sample whose magnitude exceeds a predetermined limit. The apparatus sometimes calculates and stores a final gain to replace the too-large smoothed gain (where the final gain can be obtained e.g. by dividing the to-be-replaced smoothed gain by a prior smoothed gain and multiplying the result by the corresponding prior final gain), if it is determined that the candidate gain is not less than the immediately previous final gain.

Abstract: An audio peak limiter apparatus which calculates a smoothed sequence of gains for application to a sequence of blocks of samples of an audio signal. The apparatus sometimes calculates a candidate gain, to replace a too-large smoothed gain, such that applying the candidate gain would produce no scaled sample whose magnitude exceeds a predetermined limit. The apparatus sometimes calculates and stores a final gain to replace the too-large smoothed gain (where the final gain can be obtained e.g. by dividing the to-be-replaced smoothed gain by a prior smoothed gain and multiplying the result by the corresponding prior final gain), if it is determined that the candidate gain is not less than the immediately previous final gain.

Abstract: The document relates to modulated digital filter banks, as well as to methods and systems for the design of such filter banks. In particular, the document discloses a method comprising accessing a time-domain audio signal and applying a first filter bank to the time-domain audio signal, thereby producing a first plurality of subbands of frequency-domain audio data representative of at least a part of the audio signal. The filter bank comprises an decimated modulated filter bank obtained from an asymmetric prototype filter. The method further comprises applying a second filter bank to at least a first subband of the first plurality of subbands of frequency-domain audio data, thereby producing a second plurality of subbands of frequency-domain audio data representative of at least a part of the audio signal. The second modulated filter bank comprises an asymmetric modulated filter bank which includes no decimation.

Abstract: The document relates to modulated digital filter banks, as well as to methods and systems for the design of such filter banks. In particular, the document discloses a method comprising accessing a time-domain audio signal and applying a first filter bank to the time-domain audio signal, thereby producing a first plurality of subbands of frequency-domain audio data representative of at least a part of the audio signal. The filter bank comprises an decimated modulated filter bank obtained from an asymmetric prototype filter. The method further comprises applying a second filter bank to at least a first subband of the first plurality of subbands of frequency-domain audio data, thereby producing a second plurality of subbands of frequency-domain audio data representative of at least a part of the audio signal. The second modulated filter bank comprises an asymmetric modulated filter bank which includes no decimation.