Apparatus for introducing control words by speech

• US 4,509,133 A
• Filed: 05/03/1982
• Issued: 04/02/1985
• Est. Priority Date: 05/15/1981
• Status: Expired due to Fees

• Current Assignee: Asulab SA (The Swatch Group AG)
• Original Assignee: Asulab SA (The Swatch Group AG)
• Inventors: Monbaron, Jean-Jacques, Bui, Ngoc C.
• Primary Examiner(s): Kemeny, E. S. Matt
• Application Number:
  US06/374,019
• Time in Patent Office: 1,065 Days
• Field of Search: 364/513, 364/513.5, 381/43, 381/42
• US Class Current: 704/251
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Speech recognition is improved by adaptive update of reference word codes. The apparatus can be used for controlling a piece of equipment and comprises memories which in coded form contains the references of the vocabulary of the machine; a microphone, a coding circuit and a memory for coding and memorizing the word pronounced by the user; a device for displaying the word recognized by the apparatus; and a control circuit for comparing the pronounced word to the different memorized references and for displaying at each repetition of the word the words corresponding to the references in their order of resemblance to the pronounced word. The user is able to train the apparatus easily by repeating a word until the displayed word agrees with what he has said, the coded form of this spoken word then being stored as the reference for the corresponding word of the vocabulary.

Claims

1. A device for introducing control words by speech said words being pronounced by a speaker and belonging to a predetermined vocabulary of n words, said device comprising:

means for coding each word pronounced by the speaker in accordance with a first predetermined algorithm, which gives a coded word;

means for memorizing (a) n coded references each corresponding to the coding of one of the n words of said vocabulary in accordance with the said first algorithm and (b) at least one word pronounced by said speaker in coded form;

means for displaying one of said n words;

means for initiating a change in coded references to provide a different code corresponding to a word; and

a control circuit connected to the memorizing means and to the display means, comprising;

(a) repetition detection means for comparing a presently pronounced coded word to the previous coded pronounced word, if it exists, in dependence on a second predetermined algorithm, to detect whether or not the presently pronounced word is a repetition of the previously pronounced word;

(b) means responsive to the detection means for comparing a coded pronounced word when there is no repetition, to the n references, and to allocate to each reference a level of resemblance with the coded pronounced word in dependence on a third predetermined algorithm;

(c) means for causing the display of the memorized reference having a first level of resemblance when no repetition has been detected, and for causing, if repetition has been detected, the display of the word associated with the reference occupying the level of resemblance which immediately follows that of the reference associated with the word displayed immediately previously; and

(d) means for replacing in said memorizing means the reference associated with the word displayed last, said reference being replaced by the information corresponding to the coding of the word repeated, said replacement being in response to activation by the speaker of the means for initiating a change of reference, when the speaker has found that the word displayed is the word that he actually pronounced, thereby providing a fresh coded reference for the word.

2. The device of claim 1, wherein said means for controlling a change of references comprises means for timing a delay duration d1 and means for comparing the time elapsing between a pronounced word and the previously pronounced word with the duration d1, the control means being activated when the elapsed time is greater than or equal to d1.

3. The device of claim 1, wherein said comparison means comprises means for measuring the distance between a coded pronounced word and the references;

and

means for classifying the references in accordance with an increasing order of their distances.

4. The device of claim 3, wherein said means for timing the duration d1 comprises means for producing a function f₁ (Δ

₁, Δ

₂) in which Δ

₂ represents the distance between a pronounced word and the reference occupying the second level of resemblance, and Δ

₁ represents the distance between the pronounced word and the reference occupying the first level of resemblance, the function being zero when Δ

₁ =Δ

₂ and being 1 when Δ

₁ =0 and Δ

₂ =1;

means for comparing the value of the function f₁ with a number a which is strictly between 0 and 1; and

means for giving to the duration d1 the value d_max when f₁ is less than a and the value d_max f₂ (f₁) in the opposite case, f₂ being a decreasing function of f₁ at numerical values which are strictly between 0 and 1.

5. The device of claim 3, wherein said repetition detection means comprises:

means for producing the distance δ

₁ between a pronounced word and the previously pronounced word;

means for producing the difference Δ

₁ between the pronounced word and the reference occupying the first level of resemblance;

means for producing the ratio ##EQU7## means for comparing the value of Z to a threshold S2, |S2|>

1;

means for comparing the value of δ

₁ to a first threshold S1 and a seond threshold S3 which are positive and less than 1 with S1<

S3; and

means for establishing that there is repetition if δ

₁ is less that S1 or if at the same time Z is less than S2 and δ

₁ is less than S3.

6. A speech recognition device for use with a speech controlled apparatus, said speech recognition device having a predetermined vocabulary of n words, comprising:

code means for transposing each word spoken by a speaker into a coded word;

memory means for containing (a) n codes, each code corresponding to one of the n words of the vocabulary, and (b) the code of at least one word pronounced by a speaker;

display means for displaying one of the n words; and

a control circuit connected to the display means and memory means, comprising;

(a) repetition detection means for comparing a code corresponding to the previously pronounced word to the code corresponding to the presently pronounced word to detect whether or not the presently pronounced word has been repeated;

(b) ordering means for comparing the code corresponding to the presently pronounced word, when there is no repetition, to the n codes corresponding to the n words and to allocate to each of the n codes a level of resemblance with the code of the presently pronounced word;

(c) display control means for causing the display of the word corresponding to the code having the first level of resemblance when no repetition of a word has been detected and, if a repetition has been detected, causing the display of the word corresponding to the code having a level of resemblance immediately below that of the code corresponding to the word displayed immediately previously;

(d) alteration means for replacing the code in the memory means associated with the last displayed word, said code being replaced by the code corresponding to the repeated pronounced word, said alteration means being activated by the speaker when the displayed word is the same as the pronounced word, whereby the codes in the memory are adaptively altered during the speech recognition process.