Speech recognition system including dimensionality reduction of baseband frequency signals

Claim 3

A method according to claim 1 , further comprising modelling respective ones of said demodulated plurality of baseband signals independently after the step of dimensionality reduction.

Claim 4

A method according to claim 1 , further comprising modelling respective ones of said demodulated plurality of baseband signals in a combined manner after the step of dimensionality reduction.

Claim 5

A method according to claim 1 , further wherein the step of demodulating comprises demodulating the plurality of frequency channels derived from dividing a feature extracted parameter, and wherein the step of applying dimensionality reduction comprises applying dimensionality reduction to said plurality of baseband frequency signals derived from said demodulating step.

Claim 6

A method according to claim 1 , further comprising applying a function f to said plurality of baseband signals prior to modelling, wherein said function f is given by f (N*l→M), where N*l and M are input and output dimensions and l is the feature vector length prior to channel division.

Claim 7

A method according to claim 1 , further comprising forming a feature vector by applying a transform f to {b( 1 ), d( 2 ) . . . d(n)}.

Claim 8

Apparatus for speech recognition in which a speech waveform to be modelled is represented by a set of feature extracted parameters, said apparatus comprising means for dividing individual ones of one or more of the feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to at least one of the one or more baseband frequency signals.

Claim 9

Apparatus for speech recognition in which a speech waveform to be modelled is represented by a set of feature extracted parameters, said apparatus comprising means for dividing individual ones of one or more of the feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to at least one of the one or more baseband frequency signals.

Claim 10

Apparatus according to claim 7 , further comprising means for modelling respective ones of the demodulated plurality of baseband signals independently following said means for dimensionality reduction.

Claim 11

Apparatus according to claim 7 , further comprising means for modelling respective ones of the demodulated plurality of baseband signals in a combined manner following said means for dimensionality reduction.

Claim 12

Apparatus according to claim 7 , wherein said means for demodulating comprises means for demodulating the plurality of frequency channels derived from said dividing means, and wherein said means for applying dimensionality reduction comprises means for applying dimensionality reduction to the plurality of baseband frequency signals derived from said demodulating means.

Claim 13

Apparatus according to claim 7 , further comprising means for applying a function f to the plurality of baseband signals prior to modelling wherein the function f is given by f (N*l→M), where N*l and M are input and output dimensions and l is the feature vector length prior to channel division.

Claim 14

Apparatus according to claim 7 , further comprising means for forming a feature vector by applying a transform f to {b( 1 ), d( 2 ) . . . d(n)}.

Claim 15

A speech recognition system for a radio communication system including a mobile station operating in a radio network, said speech recognition system comprising a front-end feature extraction means disposed in the mobile station, a back-end recognition means associated with the network, and apparatus for speech recognition in which a speech waveform to be modelled is represented by a set of feature extracted parameters, said apparatus comprising means for dividing individual ones of one or more of the feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to said at least one of said one or more baseband frequency signals.

Claim 16

A speech recognition system for a radio communication system including a mobile station operating in a radio network, said speech recognition system comprising a front-end feature extraction means disposed in the mobile station, a back-end recognition means associated with the network, and apparatus for speech recognition in which a speech waveform to be modelled is represented by a set of feature extracted parameters, said apparatus comprising means for dividing individual ones of one or more of the feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to said at least one of said one or more baseband frequency signals.

Claim 17

A computer program product for a speech recognition system in which a speech waveform to be modelled is represented by a set of feature extracted parameters, the computer program product comprising means for dividing individual ones of one or more of said feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to at least one of the one or more baseband frequency signals.

Claim 18

A computer program product for a speech recognition system in which a speech waveform to be modelled is represented by a set of feature extracted parameters, the computer program product comprising means for dividing individual ones of one or more of said feature extracted parameters to provide for each divided feature extracted parameter a plurality of frequency channels, means for demodulating one or more of the plurality of frequency channels to provide respective one or more corresponding baseband frequency signals, and means for applying dimensionality reduction to at least one of the one or more baseband frequency signals.

Claim 19

An article, comprising a storage medium having instructions stored thereon, the instructions when executed using a speech recognition system, in which a speech waveform to be modelled is represented by a set of feature extracted parameters, by dividing individual ones of one or more of the feature extracted parameters to provide in respect to each divided feature extracted parameter a plurality of frequency channels, demodulating at least one of the plurality of frequency channels to provide at least one corresponding baseband frequency signal, and applying dimensionality reduction to a plurality of baseband signals.

Claim 20

An article, comprising a storage medium having instructions stored thereon, the instructions when executed using a speech recognition system, in which a speech waveform to be modelled is represented by a set of feature extracted parameters, by dividing individual ones of one or more of the feature extracted parameters to provide in respect to each divided feature extracted parameter a plurality of frequency channels, demodulating at least one of the plurality of frequency channels to provide at least one corresponding baseband frequency signal, and applying dimensionality reduction to a plurality of baseband signals.