Speech recognition apparatus

Claim 2

A speech recognition apparatus comprising a microphone for converting speech into a sequence of electrical aural signals; a first processor which is provided with a first operation control circuit for computation processing to extract features of said sequence of aural signals and obtain feature patterns thereof and a first memory which is controlled by said first operation control circuit and stores said feature patterns, and which carries out feature extraction processing in sequence; a second processor which is provided with a second memory for previously storing a plurality of reference patterns and a second operation control circuit carrying out computation processing for comparing the feature patterns obtained by said first processor with said reference patterns to match each of said feature patterns to a similar reference pattern, and which carries out comparison and distinction processing in sequence; and an interface for transferring said feature patterns from the first processor to the second processor, said interface being controlled by said first and second operation control circuits, said first and second processors operating in parallel with said second processor carrying out said comparison and distinction processing for a feature pattern which has previously been extracted by said first processor while the first processor extracts features of a succeeding portion of aural signals.

Claim 3

A speech recognition apparatus as set forth in claim 1, wherein said first processor is adapted to store in said first memory said feature patterns obtained in succession with transfer of said feature patterns through said interface from the first processor to the second processor disabled, and when the second processor is ready to receive feature patterns from said first processor, said first processor transfers said feature patterns stored therein through said interface to the second processor.

Claim 4

A speech recognition apparatus as set forth in claim 1, wherein said second processor is provided with a memory in which said feature patterns transferred from said first processor are stored.

Claim 5

A speech recognition apparatus as set forth in claim 2, wherein said second processor is provided with a memory in which said feature patterns transferred from said first processor are stored.

Claim 6

A speech recognition apparatus comprising a microphone for converting speech into a sequence of electrical aural signals; a first processor carrying out feature extracting processing to extract features of said sequence of aural signals and obtain feature patterns and preliminary comparison and distinction processing in sequence for comparing said feature patterns with a plurality of reference patterns previously stored, thereby selecting a plurality of reference patterns similar to said feature patterns; a second processor carrying out final comparison and distinction processing in sequence for comparing said feature patterns obtained by said first processor with said reference patterns selected by said first processor to match said feature patterns to reference patterns which are most similar to said feature patterns; and an interface operating under control of said first and second processors for transferring said feature patterns and data related to said reference patterns selected by said first processor from said first processor to said second processor, said first and second processors operating in parallel, with said second processor carrying out said final comparison and distinction processing for the feature pattern which has previously been extracted by the first processor and the reference patterns which have previously been selected by the first processor while the first processor extracts features of a succeeding portion of aural signals and conducts preliminary comparison and distinction processing on said succeeding portion of aural signals.

Claim 7

A speech recognition apparatus comprising a microphone for converting speech into a sequence of electrical aural signals; a first processor carrying out feature extracting processing to extract features of said sequence of aural signals and obtain feature patterns and preliminary comparison and distinction processing in sequence for comparing said feature patterns with a plurality of reference patterns previously stored, thereby selecting a plurality of reference patterns similar to said feature patterns; a second processor carrying out final comparison and distinction processing in sequence for comparing said feature patterns obtained by said first processor with said reference patterns selected by said first processor to match said feature patterns to reference patterns which are most similar to said feature patterns; and an interface operating under control of said first and second processors for transferring said feature patterns and data related to said reference patterns selected by said first processor from said first processor to said second processor, said first and second processors operating in parallel, with said second processor carrying out said final comparison and distinction processing for the feature pattern which has previously been extracted by the first processor and the reference patterns which have previously been selected by the first processor while the first processor extracts features of a succeeding portion of aural signals and conducts preliminary comparison and distinction processing on said succeeding portion of aural signals.

Claim 8

A speech recognition apparatus as set forth in claim 5, wherein said first and second processors each include a memory in which said reference patterns are stored.

Claim 9

A speech recognition apparatus as set forth in claim 5, further including a memory separate from said first and second processors and in which said reference patterns are stored.

Claim 10

A speech recognition apparatus as set forth in claim 5, wherein the elapsed time required to carry out said feature extraction processing and preliminary comparison and distinction processing by said first processor and the elapsed time required to carry out said final comparison and distinction processing by said second processor are substantially equal.

Claim 11

A speech recognition apparatus as set forth in claim 6, wherein the elapsed time required to carry out said feature extraction processing and preliminary comparison and distinction processing by said first processor and the elapsed time required to carry out said final comparison and distinction processing by said second processor are substantially equal.

Claim 12

A speech recognition apparatus as set forth in claim 7, wherein the elapsed time required to carry out said feature extraction processing and preliminary comparison and distinction processing by said first processor and the elapsed time required to carry out said final comparison and distinction processing by said second processor are substantially equal.

Claim 13

A speech recognition apparatus as set forth in claim 1, wherein said comparison and distinction processing utilizes a linear matching algorithm.

Claim 14

A speech recognition apparatus as set forth in claim 1, wherein said comparison and distinction processing utilizes a dynamic time warping algorithm.

Claim 15

A speech recognition apparatus as set forth in claim 5, wherein said preliminary comparison and distinction processing utilizes a linear matching algorithm.

Claim 16

A speech recognition apparatus as set forth in claim 13, wherein said final comparison and distinction processing utilizes a dynamic time warping algorithm.

Claim 17

A speech recognition apparatus as set forth in claim 5, wherein said final comparison and distinction processing utilizes a dynamic time warping algorithm.