Speech coding apparatus for separately processing divided signal vectors

Claim 2

A speech coding apparatus receiving an input speech signal, comprising: a code book storing a N×2 M size signal vector group with 2 M indexes and a dimension of N, said code book including a number n of code book blocks, n being a positive integer greater than or equal to 2, each of the code book blocks including 2 M/n indexes and a dimension of N/n, each of the code book blocks corresponding to mutually different indexes and mutually different portions of the dimension N, and each of the code book blocks storing a 2 M/n ×N/n size group of divided signal vectors, where M and N are positive integers; a multiplier unit, coupled to said code book, which reads out signal vectors successively from said code book and selectively multiplies the signal vectors by a gain to generate an output; a prediction filter unit, coupled to said multiplier unit, which uses a given filter coefficient and performs a filter operation with respect to the output from said multiplier unit to generate a reproduced speech signal; an error detection unit, coupled to said prediction filter unit, which receives as an input the reproduced speech signal from said prediction filter unit and the input speech signal, and obtains an error between the reproduced speech and input speech signals; and an evaluation unit, operatively connected to said error detection unit, which determines the index corresponding to the signal vector giving the smallest error.

Claim 3

A speech coding apparatus receiving an input speech signal, comprising: a code book storing a N×2 M size vector group with 2 M indexes and a dimension of N, said code book including a number n of code book blocks, n being a positive integer greater than or equal to 2, each of the code book blocks including 2 M/n indexes and a dimension of N/n, each of the code book blocks corresponding to mutually different indexes and mutually different portions of the dimension N, and each of the code book blocks storing a 2 M/n ×N/n size group of divided signal vectors, where M and N are positive integers; a multiplier unit, coupled to said code book, which reads out signal vectors successively from said code book and multiplies the signal vectors by a gain to generate an output; a prediction filter unit, coupled to said multiplier unit, which uses a given filter coefficient and performs a filter operation with respect to the output from said multiplier unit to generate a reproduced speech signal; an error detection unit, coupled to said prediction filter unit, which receives as an input the reproduced speech signal from said prediction filter unit and the input speech signal and obtains an error between the reproduced speech and input speech signals; an evaluation unit, coupled to said error detection unit, which determines the index corresponding to the signal vector giving the smallest error, said multiplier unit including n number of multipliers connected to corresponding n number of code book blocks; said prediction filter unit including n number of prediction filters provided corresponding to said n number of multipliers; said error detection unit including n number of error detectors which are provided corresponding to the said n number of prediction filter units and which output n number of errors; and said evaluation unit including n number of evaluators which receive as inputs the n number of errors and determine the divided signal vectors giving the smallest corresponding errors from the corresponding code book blocks.

Claim 4

A speech coding apparatus receiving an input speech signal, comprising: a code book storing a N×2 M size vector group with 2 M indexes and a dimension of N, said code book including a number n of code book blocks, n being a positive integer greater than or equal to 2, each of the code book blocks including 2 M/n indexes and a dimension of N/n, each of the code book blocks corresponding to mutually different indexes and mutually different portions of the dimension N, and each of the code book blocks storing a 2 M/n ×N/n size group of divided signal vectors, where M and N are positive integers; a multiplier unit, coupled to said code book, which reads out signal vectors successively from said code book and multiplies the signal vectors by a gain to generate an output; a prediction filter unit, coupled to said multiplier unit, which uses a given filter coefficient and performs a filter operation with respect to the output from said multiplier unit to generate a reproduced speech signal; an error detection unit, coupled to said prediction filter unit, which receives as an input the reproduced speech signal from said prediction filter unit and the input speech signal and obtains an error between the reproduced speech and input speech signals; an evaluation unit, coupled to said error detection unit, which determines the index corresponding to the signal vector giving the smallest error, said multiplier unit including n number of multipliers connected to corresponding n number of code book blocks; said prediction filter unit including n number of prediction filters provided corresponding to said n number of multipliers; said error detection unit including n number of error detectors which are provided corresponding to the said n number of prediction filter units and which output n number of errors; and said evaluation unit including n number of evaluators which receive as inputs the n number of errors and determine the divided signal vectors giving the smallest corresponding errors from the corresponding code book blocks.

Claim 5

An apparatus as set forth in claim 2, further comprising adding means for adding a first divided signal vector giving a first smallest error specified by a first evaluator to a second code book block, the first divided signal vector being used as a fixed input and a second evaluator specifies a second divided signal vector giving a second smallest error and the specified first divided signal vector and second divided signal vector being used as fixed inputs and a third evaluator specifies a third divided signal vector giving a third smallest error, and this operation being repeated for each of the fourth and later code book blocks.

Claim 6

An apparatus as set forth in claim 3, wherein the same gain is given for all of the n number of multipliers.

Claim 7

An apparatus as set forth in claim 2, wherein when a first divided signal vector giving a first smallest error is specified by a first evaluator, the first error is fixed and in that state a second divided signal vector giving a minimum second error obtained by adding that first error to a second error detector is specified by a second evaluator, the second error is fixed and in that state a third divided signal vector giving a minimum third error obtained by adding the second error to a third error detector is specified by a third evaluator, and a similar operation is repeated for the fourth and later code book blocks.

Claim 8

An apparatus as set forth in claim 5, wherein the same gain is given to all of the n number of multipliers.

Claim 9

An apparatus as set forth in claim 2, wherein, for each of the n number of code book blocks, index information expressing n number of indexes corresponding to each of the n number of divided signal vectors respectively giving n number of smallest errors; filter coefficient information expressing the filter coefficients used in the n number of prediction filters; and multiplexing and transmission means for multiplexing and transmitting, to a speech decoding apparatus on the receiver side, gain information expressing a gain corresponding to the number of multipliers.