Voice recognition using an eigenvector

• US 4,903,306 A
• Filed: 09/29/1987
• Issued: 02/20/1990
• Est. Priority Date: 10/03/1986
• Status: Expired due to Fees

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Nakamura, Shogo
• Primary Examiner(s): NOT, DEFINED
• Assistant Examiner(s): Merecki, John A.
• Application Number:
  US07/102,559
• Time in Patent Office: 875 Days
• Field of Search: 381/42, 381/41, 381/43-45, 364/513.5
• US Class Current: 704/247
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence

Abstract

A voice pattern in the form of a matrix and comprised of a plurality of frames, each including time-spectral information and temporal information, is formed from an unknown input voice signal. The voice pattern is compared with each of the voice patterns of a library of known voices partly to select a plurality of candidate voices. Each of the library voices has a predetermined eigenvector and an inner product frequency distribution of inner products between the eigenvector and the frames of its voice pattern. Then, inner products between the voice pattern of the input voice signal and the eigenvector of each of the candidate library voice are calculated. One of the plurality of candidate library voices whose predetermined inner product frequency distribution is most similar to one of the thus calculated inner product frequency distributions is selected to identify the input voice signal.

Claims

1. A voice recognition method including the steps of:

(a) forming an input voice pattern from an input voice signal of an unknown input voice, said input voice pattern including a plurality of frames, each frame having frequency and time data;

(b) comparing less than all of said frequency and time data of said input voice pattern with a library of known voice patterns, each of said known voice patterns having its own frequency and time data, to select a plurality of candidate voice patterns, each of said candidate voice patterns having;

(1) an eigenvector having a non-zero minimum eigenvalue; and

(2) an inner product frequency distribution of inner products between (i) the frames of each of said candidate voice patterns and (ii) said eigenvector;

(c) calculating an inner product frequency distribution of inner products between the frames of said input voice pattern and the eigenvector of each of said plurality of candidate voice patterns; and

(d) selecting one of said plurality of candidate voice patterns, whose inner product frequency distribution is most similar to one of said calculated inner product frequency distributions calculated at said calculating step to thereby identify said input voice by the thus selected candidate voice pattern.

2. The method of claim 1, wherein said input voice pattern has a matrix form comprised of a first predetermined number of rows and a second predetermined number of columns, wherein each of said first predetermined number of rows constitutes a frame.

3. The method of claim 1, wherein said step of comparing is carried out by comparing a predetermined number of first frames of said input voice pattern with corresponding first frames of each of said library voice patterns.

4. The method of claim 1, wherein each of said voice patterns in said library is comprised of data constituting one word.