Mandarin speech input method for Chinese computers and a mandarin speech recognition machine

Claim 2

A speech recognition method comprising steps of: segmented a training speech syllable into an initial part and a final part; training a Continuous Hidden Markov Model (CHMM) on the initial part to create an initial part model having trained initial part model parameters; training a CHMM on the final part to create a final part model having trained final part model parameters; training a CHMM on the training speech syllable to create a syllable model using the trained initial part parameter values and the trained final part parameter values as starting parameters for the syllable model; operating on an object speech sample with the syllable model; recognizing the object speech sample as an object speech syllable based on a degree of match of the object speech sample to the syllable model; representing the object speech sample as a Chinese character in accordance with the object speech syllable.

Claim 3

A method as in claim 1 further comprising a step of training a CHMM on a Mandarin tone in an input speech syllable to create a tone model.

Claim 4

A method as in claim 2, wherein the step of training a CHMM to create a tone model includes a step of training on pitch frequency, short time energy, and duration of an input speech syllable.

Claim 5

A method as in claim 1 further comprising steps of: training a Markov Model (MM) on a sequence of chinese characters as used in context to create a Chinese language model; operating on a sequence of object speech syllables in the object speech sample with the Chinese language model; and representing the object speech sample as a Chinese character sequence in accordance with a match of the sequence of object speech syllables to the Chinese language model, thereby representing the object speech sample as a Chinese character sequence in accordance with a sequence of matches to the object speed syllables.

Claim 6

A speech recognition apparatus as in claim 4, further comprising: a display connected to the computer for illustrating a recognized Chinese syllable; and an input device connected to the computer for receiving corrections from an operator.

Claim 7

A speech recognition apparatus for Mandarin speech including high level, high rising, low dipping and high falling lexical tones, comprising: a speech signal filter for receiving a speech signal and creating a filtered analog signal; an analog-to-digital (A/D) converter connected to the speech signal filter for converting the filtered analog signal to a digital speech signal; a computer connected to the A/D converter for receiving and processing the digital signal; a pitch frequency detector connected to the computer for detecting characteristics of the pitch frequency of the speech signal thereby recognizing tones in the speech signal; a speech signal pre-processor connected to the computer for detecting the endpoints of syllables of speech signals thereby defining a beginning and ending of a syllable; a Hidden Markov Model processor connected to the computer for determining degrees of match between the speech signal and a syllable model, a tone model and a language model and recognizing speech signal syllables based on the degrees of match; a training apparatus connected to the computer for training an initial part Hidden Markov model and a final part Hidden Markov model and for training a syllable model based on trained parameters of the initial part Hidden Markov model and the final part Hidden Markov model.

Claim 8

A speech recognition apparatus for Mandarin speech including high level, high rising, low dipping and high falling lexical tones, comprising: a speech signal filter for receiving a speech signal and creating a filtered analog signal; an analog-to-digital (A/D) converter connected to the speech signal filter for converting the filtered analog signal to a digital speech signal; a computer connected to the A/D converter for receiving and processing the digital signal; a pitch frequency detector connected to the computer for detecting characteristics of the pitch frequency of the speech signal thereby recognizing tones in the speech signal; a speech signal pre-processor connected to the computer for detecting the endpoints of syllables of speech signals thereby defining a beginning and ending of a syllable; a Hidden Markov Model processor connected to the computer for determining degrees of match between the speech signal and a syllable model, a tone model and a language model and recognizing speech signal syllables based on the degrees of match; a training apparatus connected to the computer for training an initial part Hidden Markov model and a final part Hidden Markov model and for training a syllable model based on trained parameters of the initial part Hidden Markov model and the final part Hidden Markov model.

Claim 9

A speech recognition apparatus as in claim 6 further comprising: a Chinese Language Model training apparatus connected to the computer for training a Markov Chinese Language Model.

Claim 10

A speech recognition apparatus as in claim 6 further comprising: a tone training apparatus connected to the computer for training the tone model.

Claim 11

A Mandarin tone recognition method comprising steps of: dividing training syllable utterances into five groups according to tones of the syllable utterances; training a Hidden Markov Model (HMM) on the five groups of training syllable utterances to create a Mandarin tone model; operating on an object speech sample with the Mandarin tone model; recognizing the object speech sample as an object Mandarin tone based on a degree of match of the object speech sample to the Mandarin tone model; representing the object speech sample as a Mandarin tone in accordance with the object Mandarin tone.

Claim 12

A Mandarin tone recognition method comprising steps of: dividing training syllable utterances into five groups according to tones of the syllable utterances; training a Hidden Markov Model (HMM) on the five groups of training syllable utterances to create a Mandarin tone model; operating on an object speech sample with the Mandarin tone model; recognizing the object speech sample as an object Mandarin tone based on a degree of match of the object speech sample to the Mandarin tone model; representing the object speech sample as a Mandarin tone in accordance with the object Mandarin tone.

Claim 13

A method as in claim 9, wherein the step of training a HMM to create the Mandarin tone models includes a step of training on pitch frequency, short time energy, and duration of an input speech syllable.