Speech coder and speech decoder

Abstract

A speech coder includes a seed storage that stores a plurality of seeds used as an initial state of oscillation. An oscillator generates different vector sequences in accordance with values of the seeds stored in the seed storage and outputs the vector sequences as excitation vectors. A linear predictive coding synthesis filter receives, as input, the excitation vectors synthesizes the excitation vectors, and outputs a synthesized speech. The seed storage stores the plurality of seeds prepared in advance as the initial state of oscillation such that the vector sequences generated in the oscillator serve as effective excitation vectors from which the synthesized speech can be generated when the vector sequences are input to the linear predictive coding synthesis filter the oscillator receives, the seeds from the seed storage, generates, using the input seeds, vector sequences that serve as the effective excitation vectors from which the synthesized speech can be generated in the linear predictive coding synthesis filter, and outputs the vector sequences.

1 . (canceled) 
   
   
       2 . A speech coder that is a code excited linear prediction type speech coder, the speech coder comprising:
 a seed storage that stores a plurality of seeds used as an initial state of oscillation;   an oscillator that generates different vector sequences in accordance with values of the seeds stored in the seed storage and outputs the vector sequences as excitation vectors; and   a linear predictive coding synthesis filter that receives as input, the excitation vectors which are the vector sequences generated in accordance with the values of the seeds, synthesizes the excitation vectors, and outputs a synthesized speech, wherein:   the seed storage stores the plurality of seeds prepared in advance as the initial state of oscillation such that the vector sequences generated in the oscillator serve as effective excitation vectors from which the synthesized speech can be generated when the vector sequences are input to the linear predictive coding synthesis filter; and   the oscillator receives as input, the seeds from the seed storage, generates, using the input seeds, vector sequences that serve as the effective excitation vectors from which the synthesized speech can be generated in the linear predictive coding synthesis filter, and outputs the vector sequences.   
   
   
       3 . The speech coder according to  claim 2 , wherein the oscillator comprises a non-linear oscillator. 
   
   
       4 . The speech coder according to  claim 3 , wherein the non-linear oscillator comprises a non-linear digital filter. 
   
   
       5 . The speech coder according to  claim 4 , wherein the non-linear digital filter comprises a digital filter with a recursive structure in which the input vectors are zero sequences, the non-linear digital filter comprising a multiplier that multiplies a filter state by a gain, receives as input, an initial value of the filter state from the seed storage and fixes a coefficient of the multiplier such that poles lie outside a unit circuit on a Z plane. 
   
   
       6 . The speech coder according to  claim 5 , wherein a non-linear characteristic of the non-linear digital filter is produced by a complement addition characteristic. 
   
   
       7 . A speech decoder that is a code excited linear prediction type speech coder, the speech decoder comprising:
 a seed storage that stores a plurality of seeds used as an initial state of oscillation;   an oscillator that generates different vector sequences in accordance with values of the seeds stored in the seed storage and outputs the vector sequences as excitation vectors; and   a linear predictive coding synthesis filter that receives, as input, the excitation vectors which are the vector sequences generated in accordance with the values of the seeds, synthesizes the excitation vectors, and outputs a synthesized speech, wherein:   the seed storage stores the plurality of seeds prepared in advance as the initial state of oscillation such that the vector sequences generated in the oscillator serve as effective excitation vectors from which the synthesized speech can be generated when the vector sequences are input to the linear predictive coding synthesis filter; and   the oscillator receives, as input, the seeds from the seed storage, generates, using the input seeds, vector sequences that serve as the effective excitation vectors from which the synthesized speech can be generated in the linear predictive coding synthesis filter, and outputs the vector sequences.   
   
   
       8 . The speech decoder according to  claim 7 , wherein the oscillator is a non-linear oscillator. 
   
   
       9 . The speech decoder according to  claim 8 , wherein the non-linear oscillator is a non-linear digital filter. 
   
   
       10 . The speech decoder according to  claim 9 , wherein the non-linear digital filter comprises a digital filter with a recursive structure in which the input vectors are zero sequences, the non-linear digital filter comprises a multiplier that multiplies a filter state by a gain, receives, as input, an initial value of the filter state from the seed storage and fixes a coefficient of the multiplier such that poles lie outside a unit circuit on a Z plane. 
   
   
       11 . The speech decoder according to  claim 10 , wherein a non-linear characteristic of the non-linear digital filter is produced by a complement addition characteristic.