Digital speech communication system for minimizing quantizing noise

A digital speech communication system comprising a speech signal source;

• means connected to said speech signal source for partitioning the short-term spectrum of said speech signal into a plurality of selected preassigned intelligibility subband portions including means for generating a first signal corresponding to the product of said speech signal and a cosine wave of frequency equal to the midband frequency of said selected subband, means for generating a second signal corresponding to the product of said speech signal and a sine-wave of frequency equal to the midband frequency of said selected subband, first low-pass filtering means having a predetermined impulse response for producing a first convolution signal representative of the convolution of said first signal with said predetermined impulse response, and second low-pass filtering means having said predetermined impulse response for producing a second convolution signal representative of the convolution of said second signal with said predetermined impulse response;
• a plurality of subband encoding means, each subband encoding means comprising means for sampling said first convolution signal at a rate corresponding to at least twice the bandwidth of said selected subband, means for sampling said second convolution signal at a rate corresponding to at least twice the bandwidth of said selected subband, means for interleaving each sample of said sampled first convolution signal with each sample of said sampled second convolution signal in pairs, and means responsive to each pair of interleaved samples for generating first and second successive digital codes representative of said pair of interleaved samples;
• and means for combining the digital codes of said plurality of subband encoding means and for applying said combined codes to a common communication channel.

A digital speech signal communication system of predetermined intelligibility comprising means for sampling an incoming speech signal at a rate corresponding to at least twice the bandwidth of said speech signal;

• a plurality of bandpass filters each adapted to transmit a preassigned intelligibility contributing subband portion of the short-term spectrum of said speech signal, each subband lower frequency limit being an integer multiple n of the bandwidth of said subband, each subband upper frequency limit being an integer multiple n + 1 of the bandwidth of said subband;
• a plurality of subband encoding means each responsive to a selected subband portion of said short-term speech signal spectrum for producing a digital signal, each subband encoding means comprising means for sampling the selected subband portion of said speech signal at a rate equal to twice the bandwidth of said subband and means for generating an adaptive digital code representative of each sample;
• and means for combining the plurality of digital signals from said subband encoding means and for applying said combined digital signals to a communication channel;
• and speech reconstruction means including means connected to said communication channel for separating said combined digital signals into said plurality of selected subband digital signals, a plurality of subband decoding means each comprising means for generating a sequence of samples corresponding to a selected subband digital signal, means for generating a sampled subband signal from said selected subband sequence, and means responsive solely to said sampled subband signal for producing an analog replica of said selected subband portion of the speech signal, and means for combining the analog replicas of said selected subband portions to reconstruct said speech signal.

A digital speech signal communication system of predetermined intelligibility comprising means for sampling an incoming speech signal at a rate corresponding to at least twice the bandwidth of said speech signal, bandpass filter means for dividing said sampled speech signal into preassigned intelligibility contributing subband portions of the short-term speech signal spectrum, means for multiplying said selected subband portion of said speech signal from said bandpass filtering means by a sampled cosine wave of frequency equal to the lower frequency limit of said selected subband, and low-pass filtering means having a predetermined impulse response for producing a sampled signal corresponding to the convolution of the sampled signal from said multiplying means with said predetermined impulse response, a plurality of subband encoding means each responsive to a selected subband portion of said short-term speech signal spectrum for producing a digital signal, each subband encoding means comprising means for generating a digital code representative of each sample from said low-pass filtering means, and means for combining the plurality of digital signals from said subband encoding means and for applying said combined digital signals to a communication channel, and speech reconstruction means including means connected to said communication channel for separating said combined digital signals into said plurality of selected subband digital signals, a plurality of subband decoding means each comprising means for generating a sequence of samples corresponding to a selected subband digital signal, means for generating a sampled subband signal from said selected subband sample sequence, and means responsive solely to said sampled subband portion for producing an analog replica of said selected subband portion of the speech signal, and means for combining the analog replicas of said selected subband portions to reconstruct said speech signal.

A digital speech signal communication system of predetermined intelligibility comprising means for sampling an incoming speech signal at a rate corresponding to at least twice the bandwidth of said speech signal, a plurality of dividing means responsive to said sampled speech signal for dividing said sampled speech signal into preassigned intelligibility contributing subband portions of the short-term speech signal spectrum, each dividing means including means for multiplying said sampled speech signal by a sampled cosine wave of frequency equal to the midband frequency of said selected subband to produce a first sampled signal, means for multiplying said sampled speech signal by a sampled sine-wave of frequency equal to the midband frequency of said selected subband to produce a second sampled signal, first low-pass filter means having a predetermined impulse response for generating a third sampled signal representative of the convolution of said first sampled signal with said predetermined impulse response, second low-pass filter means having said predetermined impulse response for generating a fourth sampled signal representative of the convolution of said second sampled signal with said predetermined impulse response, said third and fourth sampled signals being limited to the bandwidth of said selected subband, a plurality of subband encoding means for producing a digital signal, each subband encoding means comprising first decimating means for sampling said third sampled signal at a rate equal to one-half the bandwidth of the selected subband, second decimating means for sampling said fourth sampled signal at a rate corresponding to one-half the bandwidth of the selected subband, and means for interleaving the first decimating means output samples with the second decimating means output samples to form a sampled representation of said selected subband portion of the speech signal at a rate equal to the bandwidth of said selected subband, and means for generating a digital code corresponding to said selected subband sampled representation, and means for combining the plurality of digital signals from said subband encoding means and for applying said combined digital signal to a communication channel, and speech reconstruction means including means connected to said communication channel for separating said combined digital signals into said plurality of selected subband digital signals, a plurality of subband decoding means each comprising means for generating a sequence of samples corresponding to a selected subband digital signal, means for generating a sampled subband signal from said selected subband sample sequence, means responsive solely to said sampled subband signal for producing an analog replica of said subband portion of said speech signal, and means for combining the analog replicas of said selected subband portions to reconstruct said speech signal.