A low-delay code-excited linear predictive coder for speech or audio

Claim 1

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of speech-related information, including calculating an adapted gain by which each candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as producing the smallest error value; communicating information including information representing the location in the table of the selected one of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of synthesis filter coefficients, and the step of determining the set of synthesis filter coefficients includes determining them by linear predictive analysis from a speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector.

Claim 2

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of speech-related information, including calculating an adapted gain by which each candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as producing the smallest error value; communicating information including information representing the location in the table of the selected one of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of synthesis filter coefficients, and the step of determining the set of synthesis filter coefficients includes determining them by linear predictive analysis from a speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector.

Claim 5

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of: grouping said speech into frames of speech each having at least four (4) samples with each frame representing a portion of said speech: forming a target set of excitation information in response to a present one of said frames of speech; determining a set of filter coefficients in response either to a starting signal or a frame of speech which occurred prior to said present frame of speech; calculating information representing a finite impulse response filter from said set of filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table and selected in an order based on frequency of use in prior encoding of speech, in response to the finite impulse-response filter information and each of said candidate sets and said target set of excitation information, including multiplying each of said candidate sets by a gain factor, wherein a single candidate set is used in different iterations when and if multiplied by different gain factors; selecting a candidate set of excitation information and a gain factor producing an error value below a predetermined value; and communicating information including information representing the location in the table of the selected one of the candidate sets of excitation information, said communicated information not including any information regarding the set of filter coefficients.

Claim 6

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of: grouping said speech into frames of speech each having at least four (4) samples with each frame representing a portion of said speech: forming a target set of excitation information in response to a present one of said frames of speech; determining a set of filter coefficients in response either to a starting signal or a frame of speech which occurred prior to said present frame of speech; calculating information representing a finite impulse response filter from said set of filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table and selected in an order based on frequency of use in prior encoding of speech, in response to the finite impulse-response filter information and each of said candidate sets and said target set of excitation information, including multiplying each of said candidate sets by a gain factor, wherein a single candidate set is used in different iterations when and if multiplied by different gain factors; selecting a candidate set of excitation information and a gain factor producing an error value below a predetermined value; and communicating information including information representing the location in the table of the selected one of the candidate sets of excitation information, said communicated information not including any information regarding the set of filter coefficients.

Claim 13

A method of encoding speech as claimed in any of the preceding claims including associating each of the candidate sets of excitation information for storage in the table with an index that can be efficiently communicated and arranging the association of indices to limit inaccuracy in the event of a one-bit error in communicating each of said indices.

Claim 15

A method of encoding speech as claimed in any of claims 8 to 10, including the additional step of iteratively pre-selecting the plurality candidates of excitation information from among a larger plurality of such sets for storage in the table by multiplying each of the larger plurality by each of a number of candidate gain factors, testing the products for accuracy in modeling a set of training speech samples, and reducing the larger plurality to the plurality for storage in the table on the basis of the results of the testing step.

Claim 16

A method of encoding speech as claimed in any of the preceding claims, wherein the step of determining a set of filter coefficients includes determining at least twenty of them.

Claim 17

A method of decoding speech encoded by a method as claimed in any of the preceding claims comprising the steps of initializing a synthesis filter suitably matched to the synthesis filter of the encoding method; receiving a signal produced by the encoding method; decoding the received signal to yield a set of excitation information; assembling frames of output speech from a combination of the synthesis filter and the excitation information from the decoding step, each frame having a plurality of sample of the output speech; determining a set of synthesis filter coefficients when appropriate in response to at least one frame of assembled output speech; and calculating information representing a synthesis filter from said set of filter coefficients in preparation for the subsequent portion of the received signal.

Claim 18

A method of communicating, employing an encoding method as claimed in any of claims 1 to 14 and any corresponding decoding method, wherein the iteratively calculating step of the encoding method includes on a repetition cycle extending over a plurality of frames, omitting calculations for half of the candidate sets of excitation information stored in the encoding table, and the communicating step includes inserting a synchronization bit for each occurrence of the omitting step; and the decoding method includes detecting the synchronization bit.

Claim 19

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of excitation information, including calculating an adapted gain for each of said candidate sets of excitation information by which said candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as multiplied by said adapted gain as having the smallest error value; communicating information including information representing the location in the table of the selected ones of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of filter coefficients and at least a portion of information necessary to specify said adapted gain; the step of determining the set of filter coefficients includes determining them from at least one speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector; and the gain calculating step includes calculating said portion of the information necessary to specify the adapted gain at least partly from the prior adapted gain which multiplied the selected set of excitation information for a speech vector prior to the current speech vector, including multiplying said prior gain by a weighting factor which is iteration-dependent.

Claim 20

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of excitation information, including calculating an adapted gain for each of said candidate sets of excitation information by which said candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as multiplied by said adapted gain as having the smallest error value; communicating information including information representing the location in the table of the selected ones of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of filter coefficients and at least a portion of information necessary to specify said adapted gain; the step of determining the set of filter coefficients includes determining them from at least one speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector; and the gain calculating step includes calculating said portion of the information necessary to specify the adapted gain at least partly from the prior adapted gain which multiplied the selected set of excitation information for a speech vector prior to the current speech vector, including multiplying said prior gain by a weighting factor which is iteration-dependent.

Claim 22

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of excitation information, including calculating an adapted gain for each of said candidate sets of excitation information by which said candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as multiplied by said adapted gain as having the smallest error value; communicating information including information representing the location in the table of the selected ones of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of filter coefficients and at least a portion of information necessary to specify said adapted gain; and the step of determining the set of filter coefficients includes determining a set of filter coefficients for a synthesis of at least twentieth order from at least one speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector.

Claim 23

A method of encoding speech for communication to a decoder for reproduction, comprising the steps of grouping said speech into frames of speech each having a plurality of samples with each frame representing a portion of said speech; forming a target set of speech-related information in response to at least a portion of a frame of speech which is the current frame of speech; determining a set of synthesis filter coefficients in response to at least a portion of a frame of speech; calculating information representing a synthesis filter from said set of synthesis filter coefficients; iteratively calculating an error value for each of a plurality of candidate sets of excitation information stored in a table in response to the filter information and each of said candidate sets and said target set of excitation information, including calculating an adapted gain for each of said candidate sets of excitation information by which said candidate set is multiplied prior to calculating the respective error value; selecting one of said candidate sets of excitation information as multiplied by said adapted gain as having the smallest error value; communicating information including information representing the location in the table of the selected ones of said candidate sets of excitation information for reproduction of said speech for the current frame of speech, CHARACTERIZED IN THAT the forming step includes forming a target set of speech-related information in response to a portion of the current frame of speech, which portion is the current speech vector; the communicating step excludes communicating the set of filter coefficients and at least a portion of information necessary to specify said adapted gain; and the step of determining the set of filter coefficients includes determining a set of filter coefficients for a synthesis of at least twentieth order from at least one speech vector representing at least a portion of a frame of simulated decoded speech, which vector occurred prior to the current speech vector.