Variable rate vocoder

A variable rate code excited linear prediction (CELP) coder for variable rate encoding of input frames of digitized samples of an acoustical signal comprised primarily of speech and background noise, comprising:

• means for computing linear predictive coding (LPC) coefficients for each frame of a sequence of input frames of digitized samples of an acoustical signal;
• means for selecting for each frame an output data packet rate from a set of data packet rates based upon at least one of said LPC coefficients;
• means for limiting the number of bits representing said LPC coefficients to a predetermined number as determined by said selected rate;
• means for determining pitch parameters for each pitch subframe of a set of constituent pitch analysis subframes of each frame wherein the number of pitch subframes for each frame is determined by said selected rate, and wherein said pitch parameters for each pitch subframe are represented by a number of bits as determined by said selected rate;
• means for determining codebook parameters for each codebook subframe of a set of constituent codebook analysis subframes of each frame wherein the number of codebook analysis subframes for each frame is determined by said selected rate, and wherein said codebook parameters for each codebook subframe is represented by a number of bits as determined by said selected rate;
• and
• wherein for each frame a corresponding output data packet of bits representative of said LPC coefficients, and said pitch parameters and codebook parameters for each respective pitch and codebook subframe are provided at said selected rate.

A decoder for decoding frames of variable rate code excited linear prediction (CELP) encoded speech signal, said frames having been encoded according to a voice activity measurement based upon a predetermined speech model, comprising:

• means for unpacking said encoded frames in accordance with a data rate for each of said frames to provide predetermined ones of codebook index parameters, codebook gain values, pitch values and linear prediction values;
• means for providing a corresponding codebook excitation vector in accordance with said codebook index parameters;
• means responsive to said codebook excitation vector for amplifying said codebook excitation vector in accordance with said codebook gain parameters to provide a pitch residual;
• means for pitch synthesis filtering said pitch residual in accordance with said received pitch lag and gain values to provide a formant;
• and
• means for formant synthesis filtering said formant residual in accordance with said linear prediction values to provide reconstructed speech samples for said frame.

A variable rate code excited linear prediction (CELP) coder for variable rate encoding of input frames of digitized samples of an acoustical signal comprised primarily of speech and background noise, said apparatus comprising:

• a circuit for determining for each frame of a sequence input frames of digitized speech samples a corresponding frame energy value;
• a circuit for computing linear predictive coding coefficients (LPC) for each frame of a sequence of input frames of digitized samples of an acoustical signal;
• a circuit for selecting for each frame an output data packet rate from a set of data packet rates based upon said frame energy value;
• a circuit for limiting the number of bits representing said LPC coefficients to a predetermined number as determined by said selected rate;
• a circuit for determining pitch parameters for each pitch subframe of a set of constituent pitch analysis subframes of each frame wherein the number of pitch subframes for each frame is determined by said selected rate, and wherein said pitch parameters for each pitch subframe are represented by a number of bits as determined by said selected rate;
• a circuit for determining codebook parameters for each codebook subframe of a set of constituent codebook analysis subframes of each frame wherein the number of codebook analysis subframes for each frame is determined by said selected rate, and wherein said codebook parameters for each codebook subframe are represented by a number of bits as determined by said selected rate;
• and
• wherein for each frame a corresponding output data packet of bits representative of said LPC coefficients, and said pitch parameters and codebook parameters for each respective pitch and codebook subframe is provided at said selected rate.

In a decoder for decoding variable rate code excited linear prediction (CELP) coder encoded frames of digitized samples of an acoustical signal wherein said decoder further comprises a postfilter for attenuating the reconstructed acoustical signal in the frequency band between the formant frequencies of the acoustical signal, a method for enhancing the perceptual quality of reconstructed speech by compensating for the spectral tilt of the reconstructed speech introduced by the postfiltering process, comprising the steps of:

• determining a spectral tilt value of the postfiltered speech;
• determining a set of filter coefficients in accordance with said spectral tilt value;
• and
• filtering the postfiltered speech in accordance with said determined set of filter coefficients.

A method for variable rate encoding of input frames of digitized samples of an acoustical signal comprised primarily of speech and background noise in a code excited linear prediction (CELP) coder comprising the steps of:

• determining a frame energy level for each frame of a sequence input frames of digitized speech samples;
• selecting for each frame an output data packet rate from a set of data packet rates based upon said determined frame energy level;
• computing linear predictive coding coefficients (LPC) for each frame;
• limiting the number of bits representing said LPC coefficients to a predetermined number as determined by said selected rate;
• determining pitch parameters for each pitch subframe of a set of constituent pitch analysis subframes of each frame wherein the number of pitch subframes for each frame is determined by said selected rate, and wherein said pitch parameters for each pitch subframe is represented by a number of bits as determined by said selected rate;
• determining codebook parameters for each codebook subframe of a set of constituent codebook analysis subframes of each frame wherein the number of codebook analysis subframes for each frame is determined by said selected rate, and wherein said codebook parameters for each codebook subframe is represented by a number of bits as determined by said selected rate;
• and
• providing for each frame a corresponding output data packet of bits representative of said LPC coefficients, and said pitch parameters and codebook parameters for each respective pitch and codebook subframe.

A variable rate code excited linear prediction (CELP) coder for variable rate encoding of input frames of digitized samples of an acoustical signal comprised primarily of speech and background noise, comprising:

• means for determining a frame energy level for each frame of a sequence input frames of digitized samples of an acoustical signal;
• means for selecting for each frame an output data packet rate from a set of data packet rates based upon said determined frame energy level;
• means for computing linear predictive coding (LPC) coefficients for each frame;
• means for limiting the number of bits representing said LPC coefficients to a predetermined number as determined by said selected rate;
• means for determining pitch parameters for each pitch subframe of a set of constituent pitch analysis subframes of each frame wherein the number of pitch subframes for each frame is determined by said selected rate, and wherein said pitch parameters for each pitch subframe is represented by a number of bits as determined by said selected rate;
• means for determining codebook parameters for each codebook subframe of a set of constituent codebook analysis subframes of each frame wherein the number of codebook analysis subframes for each frame is determined by said selected rate, and wherein said codebook parameters for each codebook subframe is represented by a number of bits as determined by said selected rate;
• and
• wherein for each frame a corresponding output data packet of bits representative of said LPC coefficients, and said pitch parameters and codebook parameters for each respective pitch and codebook subframe is provided at said selected rate.