Speaker recognizer in which a significant part of a preselected one of input and reference patterns is pattern matched to a time normalized part of the other

• US 4,403,114 A
• Filed: 06/30/1981
• Issued: 09/06/1983
• Est. Priority Date: 07/15/1980
• Status: Expired due to Term

• Current Assignee: Nippon Electric Company Limited (NEC Corporation)
• Original Assignee: Nippon Electric Company Limited (NEC Corporation)
• Inventors: Sakoe, Hiroaki
• Primary Examiner(s): Kemeny, Emanuel S.
• Application Number:
  US06/279,277
• Time in Patent Office: 798 Days
• Field of Search: 179/1.5 B, 179/1.5 C, 179/1.5 D, 340/146.3 A, 340/146.3 FT, 340/146.3 WD, 364/513
• US Class Current: 704/252
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• CPC Class Codes: G10L 15/12 using dynamic programming techniques, e.g. dynamic time warping [DTW]

Abstract

Speaker recognition is decided by a similarity measure (D) calculated from comparing selected feature vectors among an input speech signal sequence of feature vectors (A) and a selected sequence (B) of reference vectors selected from a plurality of pre-stored reference sequences. Prior to comparison of the input and reference vector sequences, the two sequences are time normalized to align corresponding feature vectors. A significant sound specifying signal (V) including a time sequence of elementary signals is generated in synchronism with one of the input and reference sequences and indicates which feature vectors in that one of the input and reference sequences are considered to represent significant sound. The similarity measure (D) is then calculated in accordance with the comparison of those feature vectors in the one sequence which are indicated by the significant sound specifying signal as representing significant sound and the corresponding feature vectors of the other sequence.

Claims

1. A speaker recognizing system comprising:

input time sequence producing means responsive to an input speech sound, spoken by a speaker to be recognized and comprising a significant sound of a predetermined nature informative of said speaker, for producing an input time sequence of feature vectors representative of said input speech sound;

significant sound specifying means responsive to said input speech sound for producing a sound nature signal which comprises a significant sound signal specifying said significant sound;

specific time sequence producing means for producing a specific time sequence of feature vectors representative of a specific speech sound spoken by a specific speaker, said specific speech sound comprising a significant sound informative of said specific speaker;

time normalizing means for time normalizing said input time sequence and said specific time sequences relative to each other to derive first and second normalized time sequences of feature vectors from said input time sequence and said specific time sequence, respectively;

similarity measure calculating means responsive to said sound nature signal and said first and said second normalized time sequences for calculating a similarity measure between those feature vectors of said normalized time sequences of feature vectors which are selected from said first and said second normalized time sequences in compliance with said significant sound signal, respectively, said similarity measure calculating means producing a similarity measure signal representative of the calculated similarity measure; and

means responsive to said similarity measure signal for recognizing whether or not the speaker to be recognized is said specific speaker.

2. A speaker recognizing system as claimed in claim 1, wherein said specific time sequence producing means comprises:

means for storing a plurality of stored sequences of feature vectors representative of reference speech sounds spoken by a plurality of registered speakers, each reference speech sound comprising a significant sound informative of the speaker by whom said each reference speech sound is spoken; and

sequence selecting means for selecting one of said stored sequence at a time to produce the selected one of said stored sequences as said specific time sequence, said specific speaker being that one of said registered speakers by whom the reference speech sound represented by said selected one of the stored sequences is spoken.

3. A speaker recognizing system comprising:

specific time sequence producing means for producing a specific time sequence of feature vectors representative of a specific speech sound spoken by a specific speaker, said specific speech sound comprising a significant sound of a predetermined nature informative of said specific speaker;

significant sound specifying means for producing a sound nature signal which comprises a significant sound signal specifying said significant sound;

input time sequence producing means responsive to an input speech sound spoken by a speaker to be recognized and comprising a significant sound informative of the speaker to be recognized for producing an input time sequence of feature vectors representative of said input speech sound;

time normalizing means for time normalizing said input and said specific time sequences relative to each other to derive first and second normalized time sequences of feature vectors from said input and said specific time sequences, respectively, to produce said first and said second normalized time sequences;

similarity measure calculating means responsive to said sound nature signal and said first and said second normalized time sequences for calculating a similarity measure between those feature vectors of said first and second normalized time sequences of feature vectors which are selected from said first and said second normalized time sequences in compliance with said significant sound signal, respectively, said similarity measure calculating means producing a similarity measure signal representative of the calculated similarity measure; and

means responsive to said similarity measure signal for recognizing whether or not the speaker to be recognized is said specific speaker.

4. A speaker recognizing system as claimed in claim 3, wherein said specific time sequence producing means comprises:

means for storing a plurality of stored sequences of feature vectors representative of reference speech sounds spoken by a plurality of registered speakers, each reference speech sound comprising a significant sound of a predetermined nature informative of the speaker by whom said each reference speech sound is spoken; and

sequence selecting means for selecting one of said stored sequences at a time to produce the selected one of said stored sequences as said specific time sequence, said specific speaker being that one of said registered speakers by whom the reference speech sound represented by said selected one of the stored sequences is spoken;

said significant sound specifying means comprising;

means for storing a plurality of stored nature signals in one-to-one correspondence to said stored sequences, each stored nature signal comprising a second sound signal specifying the significant sound of the reference speech sound represented by the stored sequence corresponding to said each stored nature signal; and

means operatively coupled to said sequence selecting means for selecting that one of said stored nature signals which corresponds to said specific time sequence, said significant sound specifying signal being the stored sound signal of the selected one of said stored nature signals.