International Business Machines Corporation

Nuance Communications, Inc.

Dan Chazan

Ron Hoory

Zvi Kons

Meir Zibulski

Gal Ashour

Dov Ramm

Alexander Sorin

Slava Shechtman

Gilad Cohen

Ezra Silvera

Ronen Ozer

A method for processing a speech signal includes dividing the speech signal into a succession of frames, identifying one or more of the frames as click frames, and extracting phase information from the click frames. The speech signal is encoded using the phase information. Methods are also provided for modeling phase spectra of voiced frames and click frames.

Speech synthesis using complex spectral modeling

A method for tracking pitch signal, including receiving a detected pitch signal that consists of a succession of pitch values, and for each current pitch value in the detected signal perform the following steps: constructing sub-sequences of consistent pitch values from neighboring pitch values. Next, calculating significance of the sub-sequences, and selecting a sub-sequence or a collection of consistent subsequences with highest significance. If the current pitch value is not consistent with the sub-sequence with highest significance, smoothing the current pitch value by diving it or multiplying it by an integer value&#x3e;1, so as to render it consistent with the sub-sequence with highest significance.

Method for tracking a pitch signal

A speech decoder and a segment aligner are provided in the present invention. The speech decoder may include a spectrum reconstructor operative to reconstruct the spectrum of a speech segment from the amplitude envelope of the spectrum of said speech segment and pitch information, a phase combiner operative to reconstruct the complex spectrum of the speech segment from the reconstructed spectrum, phase information describing the speech segment, and pitch information describing the speech segment. The speech decoder may further include a delay operative to store a complex spectrum of a previous speech segment; and a segment aligner operative to determine the relative offset between the complex spectrum of the speech segment and the complex spectrum of the previous speech segment, align the position of the first pitch excitation of the current speech segment to the last pitch excitation of the previous speech segment; and to apply a time shift and a complex Hilbert filter to said complex spectra, wherein the segment aligner is operative to cross-correlate the complex spectra as 

Method for encoding and decoding spectral phase data for speech signals

A method for speech synthesis includes receiving an input speech signal containing a set of speech segments, and estimating spectral envelopes of the input speech signal in a succession of time intervals during each of the speech segments. The spectral envelopes are integrated over a plurality of window functions in a frequency domain so as to determine elements of feature vectors corresponding to the speech segments. An output speech signal is reconstructed by concatenating the feature vectors corresponding to a sequence of the speech segments.

Feature-domain concatenative speech synthesis

A speech reconstruction method and system for converting a series of binned spectra or functions thereof such as the Mel Frequency Cepstra Coefficients (MFCC), of an original digitized speech signal, into a reconstructed speech signal, where each binned spectrum has a respective pitch value and voicing decision. The binned spectra are derived from the original digitized speech signal at successive instances by multiplying each estimate of the spectral envelope by a predetermined set of frequency domain window functions and computing the integrals thereof. At each respective time instance, harmonic frequencies and weights are generated according to the respective pitch value and voicing decision. Basis functions having bounded supports on the frequency axis are each sampled at all said harmonic frequencies, which are within its support and multiplied by respective harmonic weights. The sampled basis functions are combined with respective phases, generated according to the pitch value, voicing decision and possibly the binned spectrum, resulting in a complex line spectrum corresponding to each basis function. Coefficients are generated of the basis functions, and each of the points of the respective complex line spectra is multiplied by the respective basis function coefficient. The complex line spectra are summed up to generate for each time instance a single complex line spectrum with values for all harmonic frequencies. A time signal is generated from complex line spectra computed at successive instances of time.

Method and system for speech reconstruction from speech recognition features, pitch and voicing with resampled basis functions providing reconstruction of the spectral envelope

A method for encoding a digitized speech signal so as to generate data capable of being decoded as speech. A digitized speech signal is first converted to a series of feature vectors using for example known Mel-frequency Cepstral coefficients (MFCC) techniques. At successive instances instance of time a respective pitch value of the digitized speech signal is computed, and successive acoustic vectors each containing the respective pitch value and feature vector are compressed so as to derive therefrom a bit stream. A suitable decoder reverses the operation so as to extract the features vectors and pitch values, thus allowing speech reproduction and playback. In addition, speech recognition is possible using the decompressed feature vectors, with no impairment of the recognition accuracy and no computational overhead.

Method and system for low bit rate speech coding with speech recognition features and pitch providing reconstruction of the spectral envelope

A method for estimating a pitch frequency of an audio signal includes computing a first transform of the signal to a frequency domain over a first time interval, and computing a second transform of the signal to the frequency domain over a second time interval, which contains the first time interval. A line spectrum of the signal is found, based on the first and second transforms, the spectrum including spectral lines having respective line amplitudes and line frequencies. A utility function that is periodic in the frequencies of the lines in the spectrum is then computed. This function is indicative, for each candidate pitch frequency in a given pitch frequency range, of a compatibility of the spectrum with the candidate pitch frequency. The pitch frequency of the speech signal is estimated responsive to the utility function.

Fast frequency-domain pitch estimation

An audio mixer system is described for producing coded output in which at least a left audio signal, a right audio signal and a surround audio signal are encoded in two output channels so that the surround signal can be decoded from the difference of the two output channels. The system comprises means for generating position data designating a desired position for a sound source in a 360 degree sound field. Logic is provided for determining the relative volume of the sound source in the left, right and surround audio signals from the position data. A signed continuity factor is maintained so that the sign of the continuity factor is changed in response the desired position crossing a nominal position of the surround signal in the sound field and logic is provided for encoding the sound source data into the two output channels in accordance with the determined relative volume of the sound source in at least two of the left, right and surround signals each multiplied by the continuity factor. This reduces audible artifacts associated with phase discontinuities in the output signals either side of the surround speaker nominal position.

Audio mixer

An adaptive noise cancellation device comprises: convolution logic (10) for convolving the signal from a reference input (x) with a discretized L-tap filter to form a filtered reference signal; and logic (20) for subtracting the filtered reference signal from a signal input to form an output signal; logic for generating the filter taps as a linear combination of N basis functions each having a corresponding coefficient C.sub.k ; and logic for repeatedly determining the coefficients C.sub.k which minimize the power in the output signal (z), characterized in that N is less than the number of filter taps L and the basis functions have respective values over a portion of finite width, outside of which portion the functions are substantially zero, both in the frequency and time domains, in an embodiment they are gaussian. A full-duplex speakerphone is disclosed including such a noise cancellation device.

Growing community of inventors

Haifa, Israel

Dan Chazan