Speech processing apparatus and methods

Claim 1

Speech processing apparatus comprising: first means for deriving a first signal in accordance with the utterances made by a speaker, second means responsive to the physical forces, including at least one of the forces of physical vibration and acceleration acting on the speaker, which have the effect of altering the characteristics of the voice sounds uttered by the speaker; storage means containing a reference vocabulary in the form of patterned templates of words; and a processing unit receiving the output from the first means and from the second means and performing dynamic adaptation on said templates in accordance with the output of said second means to thereby modify said templates to anticipate the behavior of the speaker's speech production mechanism as affected by the physical forces detected by said second means; said processing unit comparing the first output signal with the modified templates to produce a second output signal in accordance with the speech sounds uttered by the speaker which are thus compensated, at least in part, by the physical forces acting on the speaker.

Claim 7

A method of providing speech processing comprising the steps of: deriving first signals from speech sounds uttered by a speaker; deriving second signals having a manifestation representative of at least one of the environmental physical forces corresponding respectively to the acceleration and vibration forces acting on the speaker of the kind which tend to affect the nature of the speech utterances made by the speaker; performing dynamic adaptation on stored pattern templates of words in a vocabulary store in accordance with said force signals to modify the templates to anticipate the behavior of the speaker's speech production mechanism during the force; and comparing the said speech signals with the modified templates to provide thereby an indication of the similarity between speech sounds and stored words that is compensated at least in part for the forces.

Claim 8

A method of providing speech processing comprising the steps of: deriving first signals from speech sounds uttered by a speaker; deriving second signals having a manifestation representative of at least one of the environmental physical forces corresponding respectively to the acceleration and vibration forces acting on the speaker of the kind which tend to affect the nature of the speech utterances made by the speaker; performing dynamic adaptation on stored pattern templates of words in a vocabulary store in accordance with said force signals to modify the templates to anticipate the behavior of the speaker's speech production mechanism during the force; and comparing the said speech signals with the modified templates to provide thereby an indication of the similarity between speech sounds and stored words that is compensated at least in part for the forces.