Apparatus and method for recognizing words from among continuous speech

• US 3,943,295 A
• Filed: 07/17/1974
• Issued: 03/09/1976
• Est. Priority Date: 07/17/1974
• Status: Expired due to Term

• Current Assignee: Siemens Corporate Research & Support, Inc. (Siemens AG)
• Original Assignee: Threshold Technology, Inc.
• Inventors: Martin, Thomas B., Herscher, Marvin B.
• Primary Examiner(s): Robinson, Thomas A.
• Assistant Examiner(s): Robinson, Thomas A.
• Application Number:
  US05/489,231
• Time in Patent Office: 601 Days
• Field of Search: 179/15 BY, 179/15 BS, 179/1 SB, 179/1 SA, 178/69.5 R, 340/146.3 Q
• US Class Current: 704/253
• CPC Class Codes: G10L 15/12 using dynamic programming techniques, e.g. dynamic time warping [DTW]

Abstract

An apparatus and method for recognizing the occurrence of a command word from among input speech that may be continuous in form. In accordance with the invention, means are provided for generating feature signals which depend on the features of an input word being spoken. The feature signals are processed to determine the time interval of occurrence of a predetermined sequence of features. Also provided is a means for comparing the feature signals which occur during the determined time interval with a stored set of features that are expected to occur characteristically during the command word to determine the degree of correlation therebetween.

Claims

1. Apparatus for recognizing the occurrence of a command word from among input speech that may be of continuous form, comprising:

a. means for generating feature signals which depend on the features of an input word being spoken;

b. means for sequentially processing the feature signals to determine the time interval of occurrence of a predetermined sequence of features; and

c. means for comparing the feature signals which occur during said time interval with a stored set of features that are expected to occur characteristically during the command word to determine the degree of correlation therebetween.

2. Apparatus for recognizing the occurrence of a command word from among input speech that may be continuous in form, comprising:

a. feature extraction means for processing the input received speech and generating feature output signals on particular ones of a number of feature output lines, the particular ones depending on the features present in an input word being spoken;

b. a sequential logic chain which includes a plurality of logic units having logic input terminals, the logic units being sequentially activated by the presence of signals on logic input terminals;

c. means for coupling selected ones of the feature output lines to the logic input terminals, the coupling selections depending on the characteristic features expected in the command word;

d. means for storing, as a time dependent matrix, the status of the feature signals which are expected to occur characteristically during the command word;

e. means for forming a time dependent matrix whose members reflect the status of feature signals that occur during a time interval determined by activation of said sequential logic chain; and

f. means for comparing the formed matrix with the stored matrix to determine the degree of correlation therebetween.

3. A method of recognizing the occurrence of a command word from among received speech that may be of continuous form, comprising the steps of:

a. generating feature signals which depend on the characteristic features of a word being spoken;

b. sequentially processing the feature signals to determine the time interval of occurrence of a predetermined sequence of features; and

c. comparing the feature signals which occur during said time interval with a stored set of features that are expected to occur during the command word to determine the degree of correlation therebetween.

4. Apparatus for recognizing the occurrence of a plurality of preselected command words from among input speech that may be continuous in form, comprising:

a. means for generating feature signals which depend on the characteristic features of a word being spoken;

b. a plurality of sequential logic circuits, each of which corresponds to a different one of the preselected command words and processes the feature signals to determine the time interval of occurrence of the predetermined sequence of features associated with its characteristic command word;

c. means for storing a set of features for each preselected command word, each set containing those features that are expected to occur during the set'"'"'s associated command word; and

d. means for comparing the feature signals which occur during the time interval measured by a particular one of the sequential logic circuits with the stored set of features for the command word corresponding to the particular sequential logic circuit.

5. Apparatus for recognizing the occurrence of a plurality of preselected command words from among input speech that may be continuous in form, comprising:

a. feature extraction means for processing the input speech and generating feature output signals on particular ones of a number of feature output lines, the particular ones depending on the features present in an input word being spoken;

b. a plurality of sequential logic chains, each of which corresponds to a different one of the preselected command words and each of which includes a plurality of logic units having logic input terminals, the logic units being sequentially activated by the presence of signals on logic input terminals;

c. means for coupling selected ones of the feature output lines to the logic input terminals, the coupling selections for each logic chain depending on the characteristic features expected in its corresponding command word;

d. means for storing, as time dependent matrices, the status of feature signals which are expected to occur characteristically during each command word;

e. means for forming a time dependent matrix whose members reflect the status of feature signals that occur during the time corresponding to activation of a particular one of said sequential logic chains; and

f. means for comparing the formed matrix with the stored matrix for the particular command word to determine the degree of correlation therebetween.

6. Apparatus as defined by claim 5 further comprising feature memory means for storing the feature output signals from said feature extraction means during operation of said sequential chains and for selectively reading out the feature signals which occurred during activation of a particular sequential logic chain.

7. Apparatus as defined by claim 6 further comprising means for recording the active time each sequential logic chain and for controlling readout of said feature memory means in accordance with said active time.