Codebook structure for changeable pulse multimode speech coding

A speech coding system comprising:

• speech processing circuitry disposed to receive a speech waveform;
• where the speech processing circuitry comprises a codebook having a plurality of subcodebooks with at least two different subcodebooks,
• where each subcodebook comprises a plurality of pulse locations for generation of at least one codevector in response to the speech waveform,
• where the plurality of pulse locations comprises at least one track, and where the at least one codevector comprises at least one pulse selected from the at least one track,
• where the at least one pulse comprises a first pulse and a second pulse,
• where the at least one track comprises a first track and a second track, and
• where the first pulse is selected from the first track and the second pulse is selected from the second track.

A speech coding system comprising:

• speech processing circuitry disposed to receive a speech waveform;
• where the speech processing circuitry comprises a codebook having a plurality of subcodebooks with at least two different subcodebooks,
• where each subcodebook comprises a plurality of pulse locations for generation of at least one codevector in response to the speech waveform, and
• where the plurality of subcodebooks comprises:
• a first subcodebook to provide a first codevector comprising a first pulse and a second pulse;
• a second subcodebook to provide a second codevector comprising a third pulse, a fourth pulse, and a fifth pulse;
• and
• a third subcodebook to provide a third codevector comprising a sixth pulse, a seventh pulse, an eighth pulse, a ninth pulse, and a tenth pulse.

A speech coding system comprising:

• speech processing circuitry disposed to receive a speech waveform;
• where the speech processing circuitry comprises a codebook having a plurality of subcodebooks with at least two different subcodebooks,
• where each subcodebook comprises a plurality of pulse locations for generation of at least one codevector in response to the speech waveform, and
• where the plurality of subcodebooks further comprises:
• a first subcodebook to provide a first codevector comprising a first pulse and a second pulse;
• and
• a second subcodebook to provide a second codevector comprising a third pulse, a fourth pulse, and a fifth pulse.

A method of searching for a codevector in a speech coding system having at least one of a pulse codebook and a pulse subcodebook, the codevector responsive to a speech waveform and having at least two pulses, the method comprising:

• conducting a first search turn for a candidate codevector;
• calculating a first criterion value in response to a location, a sign and a magnitude for each pulse resulting from said conducting the first search turn;
• conducting at least one additional search turn for at least one additional candidate codevector;
• calculating at least one additional criterion value in response to a location, a sign, and a magnitude of each pulse resulting from the at least one additional search turn;
• and
• selecting the codevector in response to the first criterion value and the at least one additional criterion value.

A method of searching for a codevector in a speech coding system having at least one pulse codebook or pulse subcodebook with a plurality of codevectors, each codevector having at least three pulses, where each pulse has a location, sign, and magnitude, and where different combinations of the pulses are different codevectors, the method comprising;

• jointly selecting locations, signs and magnitudes of a first two pulses (P1, P2);
• jointly selecting locations, signs and magnitudes of a next two pulses (Pi, Pi+1);
• until
• jointly selecting locations, signs and magnitudes of a last two pulses (PN−1, PN);
• selecting a combination of the pulses as a candidate codevector;
• and
• sequentially searching in at least two search turns from a first pair of pulses to a last pair of pulses, where a next search turn yields a smaller error signal than a previous search turn.