Continental Automotive Systems Corporation

Motorola Corporation

Motorola Mobility LLC

Google Inc.

Jianming Song

John C Johnson

Guillaume Lamy

Yaxin Zhang

Anton Madievski

Joel A Clark

David L Barron

Bijal Joshi

Mike Reuter

Ho Chuen Choi

Xiaoyuan Zhu

Approaches for detecting and reducing wind noise from audio signals captured at multi-microphone array are described. In aspects, the wind noise detector is constructed from probabilities of speech presence and wind noise presence, which are derives from statistics of the phase differences among the time aligned signals of multi-microphones in separate frequency regions. Wind noise, if detected, is reduced by a gain in frequency domain, which is also a function of the phase difference and its statistics.

Method and apparatus for wind noise attenuation

Speech in a motor vehicle is improved by suppressing transient, 'non-stationary' noise using pattern matching. Pre-stored sets of linear predictive coefficients are compared to LPC coefficients of a noise signal. The pre-stored LPC coefficient set that is 'closest' to an LPC coefficient set representing a signal comprising speech and noise is considered to be noise.

Robust noise estimation for speech enhancement in variable noise conditions

A noise suppression circuit for use in an audio signal processing circuit is provided. The noise suppression circuit includes a plurality of different types of noise activity detectors, which are each adapted for detecting the presence of a different type of noise in a received signal. The noise suppression circuit further includes a plurality of different types of noise reduction circuits, which are each adapted for removing a different type of detected noise, where each noise reduction circuit respectively corresponds to one of the plurality of noise activity detectors. The respective noise reduction circuit is then selectively activated to condition the received signal to reduce the amount of the detected types of noise, when each one of the plurality of noise activity detectors detects the presence of a corresponding type of noise in the received signal.

Method and noise suppression circuit incorporating a plurality of noise suppression techniques

Acoustic noise in an audio signal is reduced by calculating a speech probability presence (SPP) factor using minimum mean square error (MMSE). The SPP factor, which has a value typically ranging between zero and one, is modified or warped responsive to a value obtained from the evaluation of a sigmoid function, the shape of which is determined by a signal-to-noise ratio (SNR), which is obtained by an evaluation of the signal energy and noise energy output from a microphone over time.

Speech probability presence modifier improving log-MMSE based noise suppression performance

Speech from a driver and speech from a passenger in a vehicle is selected directionally using a plurality of directional microphones. Sounds detected as coming from a passenger from a plurality of directional microphones are suppressed from sounds detected as coming from a driver by a second plurality of directional microphones.

Cotalker nulling based on multi super directional beamformer

A continuous stream of noise is created from a plurality of input signals. A smoothing spectrum estimate is continuously calculated from the continuous stream of noise. Noise is responsively removed from a selected one of the plurality of input signals using the smoothing spectrum estimate. The removal of the noise from the selected input signal is performed substantially synchronously and in time alignment with the creating of the continuous stream of noise and the calculating of the smoothing spectrum estimate.

Apparatus and method for noise removal by spectral smoothing

A method () and apparatus () seeks to improve the intelligibility of speech emitted into a noisy environment. Formants are identified () and perceptual frequency scale band is selected () that includes at least one of the identified formants. The SNR in each band is compared () to a threshold and, if the SNR for that band is less than the threshold, the method increases a formant enhancement gain for that band. A set of high pass filter gains () is combined () with the formant enhancement gains yielding combined gains that are then clipped (), scaled () according to a total SNR, normalized (), smoothed across time () and frequency (), and used to reconstruct () an audio signal.

Method and apparatus for increasing speech intelligibility in noisy environments

A method () and apparatus () for enhancing the intelligibility of speech emitted into a noisy environment. After filtering () ambient noise with a filter () that simulates the physical blocking of noise by a at least a part of a voice communication device () a frequency dependent SNR of received voice audio relative to ambient noise is computed () on a perceptual (e.g. Bark) frequency scale. Formants are identified () and the SNR in bands including certain formants are modified () with formant enhancement gain factors in order to improve intelligibility. A set of high pass filter gains () is combined () with the formant enhancement gains factors yielding combined gains which are clipped (), scaled () according to a total SNR, normalized (), smoothed across time () and frequency () and used to reconstruct () an audio signal.

Method and apparatus of increasing speech intelligibility in noisy environments

A device () performs statistical pattern recognition using model parameters that are refined by optimizing an objective function that includes a term for many items of training data for which recognition errors occur wherein each term depends on a relative magnitude of a first score for a recognition result for an item of training data and a second score calculated by evaluating a statistical pattern recognition model identified by a transcribed identity of the training data item with feature vectors extracted from the item of training data. The objective function does not include terms for items of training data for which there is a gross discrepancy between a transcribed identity and a recognized identity. Gross discrepancies can be detected by probability score or pattern identity comparisons. Terms, of the objective function are weighted based on the type of recognition error and weights can be increased for high priority patterns.

Method of refining statistical pattern recognition models and statistical pattern recognizers

A voiced/unvoiced speech classifier ( ) includes a speech segmentor ( ) which segments an input digitized speech waveform into frames of speech and a band-pass filter ( ) which filters the frames of speech. A relative energy generator ( ) generates a relative energy value for each filtered frame of speech and a decision parameter generator ( ) including an autocorrelation calculator ( ) and a pitch calculator ( ) generates a decision parameter based on an autocorrelation function and a pitch frequency index for the filtered frames of speech. A normalized energy calculator ( ) adjusts the threshold and then normalizes the relative energy. A comparator ( ) provides a signal indicative of whether a frame of speech is voiced speech or unvoiced speech depending on a comparison of the decision parameter and the normalized relative energy value for each filtered frame of speech.

Voiced/unvoiced speech classifier

A method and apparatus for speech recognition involves classifying ( ) a digitized speech segment according to whether the speech segment comprises voiced or unvoiced speech and utilizing that classification to generate tonal feature vectors ( ) of the speech segment when the speech is voiced. The tonal feature vectors are then combined ( ) with other non-tonal feature vectors ( ) to provide speech feature vectors. The speech feature vectors are compared ( ) with previously stored models of speech feature vectors ( ) for different segments of speech to determine which previously stored model is a most likely match for the segment to be recognized.

Tone based speech recognition

An apparatus for selecting a cohort model for use in a speaker verification system includes a model generator ( ) for determining a target speaker model ( ) from a speech sample collected from the target speaker ( ). A cohort selector ( ) determines a similarity value between each of a number of predetermined existing speaker models from a model pool ( ) and the target speaker model ( ) and a dissimilarity value between each of the existing speaker models and any previously selected cohort models ( ). An existing speaker model which is most similar to the target speaker model, but most dissimilar to previously chosen cohort models, is then chosen as another cohort model for the target speaker.

Growing community of inventors

Barrington, IL, United States of America

Jianming Song