Method for operating a speech recognition system

• US 20060200345A1
• Filed: 10/24/2003
• Published: 09/07/2006
• Est. Priority Date: 11/02/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Koninklijke Philips Electronics N.V. (Koninklijke Philips N.V.)
• Inventors: Kooiman, Albert
• Granted Patent: US 8,781,826 B2
• Time in Patent Office: Days
• US Class Current: 704/233
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence
• CPC Class Codes: G10L 2015/225 Feedback of the input speech
• CPC Class Codes: G10L 2015/226 using non-speech characteristics

Abstract

A method for operating a speech recognition system is described in which a speech signal (S₁) of a user is detected and analyzed so as to recognize speech information contained in the speech signal (S). The speech recognition system determines a reception quality value (S_Q) or a noise value which represents a current reception quality. The speech recognition system is switched over to a mode of operation which is less sensitive to noise and/or outputs an alert signal (S_W) to the user when the reception quality value (S_Q) drops below a given reception quality threshold or when the noise value exceeds a noise threshold. An appropriate speech recognition system is also described.

Claims

1. A method for operating a speech recognition system (1), in which method a speech signal (S_I) of a user is detected and analyzed so as to recognize speech information contained in the speech signal (S_I), characterized in that there is determined a reception quality value (S_Q) or a noise value which represents a current reception quality, and in that the speech recognition system (1) switches over to a mode of operation which is less sensitive to noise and/or outputs an alert signal (S_W) to the user when the reception quality value (S_Q) drops below a given reception quality threshold or when the noise value exceeds a noise threshold.

2. A method as claimed in claim 1, characterized in that the speech recognition system is automatically reset to the previous mode of operation when the reception quality value (S_Q) exceeds the reception quality threshold again or when the noise value drops below the noise threshold again.

3. A method as claimed in claim 1, characterized in that when the reception quality value drops below the reception quality threshold or the noise value exceeds the noise threshold, a barge-in mode of operation of the speech recognition system (1) is deactivated.

4. A method as claimed in claim 1, characterized in that the reception quality value (S_Q) or the noise value is determined by means of a voice activity detector (5).

5. A method as claimed in claim 1, characterized in that the reception quality value (S_Q) or the noise value is determined on the basis of a background signal which is received prior to the beginning of the utterance and/or in a speech pause of the user.

6. A method as claimed in claim 4, characterized in that the voice activity detector (5) applies the reception quality value (S_Q) or the noise value itself and/or, when the reception quality value drops below the reception quality threshold or when the noise value exceeds the noise threshold, a reception corruption indication signal (S_EB) to a dialog control device (10).

7. A method as claimed in claim 1, characterized in that when the reception corruption indication signal (S_EB) is received and/or when the received reception quality value (S_Q) drops below the reception quality threshold or the noise value exceeds the noise threshold, the dialog control device (10) initiates the output of a prompt (S_W) to the user who is thus given the information that the reception conditions are poor.

8. A method as claimed in claim 1, characterized in that an incoming signal is analyzed as regards the type of disturbance causing the reception quality value (S_Q) to be below the reception quality threshold or the noise value to be above the noise threshold, and that a prompt (S_W) which contains this information is output to the user.

10. A computer program which comprises program code means for carrying out all steps of a method as claimed in claim 1 when the program is run on a computer.

9. A speech recognition system (1) which comprises means (5) for the detection of a speech signal (S_I) of a user and a speech recognition device (7) for analyzing the detected speech signal (S_I) so as to recognize speech information contained in the speech signal, characterized in that it comprises a quality control device (6) for determining a reception quality value (S_Q) or a noise value, representing a current reception quality,

a comparator for comparing the reception quality value (S_Q) with a predetermined reception quality threshold or for comparing the noise value with a given noise threshold,

and control means (9, 10) which are constructed in such a manner that the speech recognition system (1) is switched over to a mode of operation which is less sensitive to noise and/or an alert signal (S_W) is output to the user when the reception quality value drops below the reception quality threshold or when the noise value exceeds the noise threshold.