Microphone initialization enhancement for speech recognition

Abstract

A method and arrangement for improved speech recognition in a telephonically challenging speakerphone in-car environment. The method includes receiving a signal from a microphone representative of speech to be recognised, performing detection of a transition in the signal indicative of switch on of the microphone, and, in response to the detection, performing speech recognition on the signal with reduced contribution from an initial portion thereof. The initial portion may be treated as optional speech, the speech recognition may be performed with a predetermined redundant sound, and a user may be requested to speak the predetermined redundant sound when speech recognition has fallen below a predetermined threshold. Thus, recognition may be made possible when otherwise it would not be possible, recognition match scoring will be increased as the low weighting given by deleted initial sounds will be eliminated and therefore confusion of the recognised phrase will be reduced.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of United Kingdom Patent Application No. 0414711.2 filed Jul. 1, 2004.

FIELD OF THE INVENTION

This invention relates to speech recognition, and more particularly to a microphone initialization enhancement for automated speech recognition systems.

BACKGROUND OF THE INVENTION

In the field of this invention it is known that many speaker phones and mobile phone car installations suppress the microphone while sound is being played out in order to eliminate echo and feedback. According to conventional teachings, the microphone remains in an off state until a relatively high volume audio signal is received. This microphone enablement technique causes the microphone to remain in an off state when a user speaks in a low to medium volume. Because the speaker phone or mobile kit does not react until after a high energy audio signal is received, any low level sounds at the start of the utterance can be lost. Accordingly, the system only receives a truncated part of the utterance resulting in a speech recognition of the utterance being unsuccessful.

This approach has the disadvantage that the spoken phrase is highly likely to be rejected outright or possibly confused with other candidate phrases. A need therefore exists for method of increasing speech recognition performance in such systems wherein the abovementioned disadvantage(s) may be alleviated.

SUMMARY OF THE INVENTION

In accordance with a first aspect of the present invention there is provided a method of speech recognition for use with a system having a microphone. The method includes receiving a signal from the microphone representative of speech to be recognized, performing detection of a transition in the signal indicative of switch on of the microphone, and, in response to the detection, performing speech recognition on the signal with reduced contribution from an initial portion thereof.

In accordance with a second aspect of the present invention there is provided a speech recognition system for use with a system having a microphone. The system includes means for receiving a signal from the microphone representative of speech to be recognized, means for performing detection of a transition in the signal indicative of switch on of the microphone, and means for performing, in response to the detection, speech recognition on the signal with reduced contribution from an initial portion thereof.

It should be noted that various aspects of the invention can be implemented as a program for controlling computing equipment to implement the functions described herein, or a program for enabling computing equipment to perform processes corresponding to the steps disclosed herein. This program may be provided by storing the program in a magnetic disk, an optical disk, a semiconductor memory, any other recording medium, or can also be provided as a digitally encoded signal conveyed via a carrier wave. The described program can be a single program or can be implemented as multiple subprograms, each of which interact within a single computing device or interact in a distributed fashion across a network space.

BRIEF DESCRIPTION OF THE DRAWINGS

One method of improved speech recognition with speaker phones and car kits incorporating the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

FIG. 1 shows a schematic illustration of a prior art speakerphone speech recognition system;

FIG. 2 shows a block schematic diagram depicting a speakerphone speech recognition system incorporating the present invention; and

FIG. 3 shows a schematic illustration of a speakerphone speech recognition system.

DESCRIPTION OF PREFERRED EMBODIMENTS

Referring firstly to FIG. 1, in a known speech recognition system (not shown), the system receives a waveform 10 resulting from speaking of the words “Balance of account” 20. Because, as discussed above, the system microphone (not shown) is typically suppressed while sound is being played out in order to eliminate echo and feedback, the system does not react until after a high energy audio signal is received; therefore, sounds at the beginning of the utterance are lost. Thus, the recognition system hears “lance of account” which does not contain all of the sounds expected to be heard in “Balance of account”. Typically, such a speech recognition system is looking for a specific sequence of sounds and will either not match the expected phrase or will score the match with a low probability.

Referring now to FIG. 2, a telephone based automated speech recognition system for use by a user 100 in challenging telephony environments (such as in-car) includes a speakerphone arrangement 200 having a controller 201 which controls operation of a speaker 202 and microphone 203. A speech recognition arrangement 300 is coupled to the speakerphone arrangement 200 and has a speech detector 301, a speech recognition controller 302 and a speech recognition engine 303.

The speech detector 301 serves to automatically detect and enable the following mode of operation: the speech detector 301 identifies if the profile of the audio energy received from the speakerphone arrangement 200 is silence or close to silence followed by a high energy edge (rapid transition caused by the microphone 203 switching on); then, this situation will be notified by the speech recognition controller 302 to the speech recognition engine 303.

The speech recognition engine 303 will process the utterance but will, in light of the signal from the speech detector 301, automatically modify its behavior to not expect to match the initial sounds of the utterance. In effect, the initial sounds will become an optional part of the utterance.

Another example is to say that the recognition engine will accept deletions of sounds at the start of the utterance. Exactly how much (time) of the utterance might be allowed to be missing may be established by tuning the system for optimum recognition.

Alternatively, the grammars and call flows of the speech recognition engine may be constructed to optionally accept a redundant word (such as “please”) at the start of every utterance. This word is defined with special purpose “sound sequences” for this method which allow all possible tail ends of the utterance to match.

When the system detects repeated failures, it suggests to the user that they use the “microphone enabling” word which has the effect that the first thing the user says to interrupt the system is not required to match.

Good recognition can proceed with the significant portion of the grammar after this redundant word.

The core of this alternative mode of operation is illustrated with reference to FIG. 3. As shown in FIG. 3, the speech recognition arrangement 300 receives a waveform 410 resulting from speaking of the words “Please balance of account”. Because, as discussed above, the microphone 203 is typically suppressed while sound is being played out in order to eliminate echo and feedback, the system does not react until a high energy audio signal is received. Thus, the recognition system receives “ease balance of account”, but it is expecting phrases such as:

• • “Please balance of account” or
  • “lease balance of account” or
  • “ease balance of account”—(Match) or
  • “se balance of account” or
  • “balance of account” Therefore the desired phrase is successfully recognised.

It will be understood that this method can be implemented without modification to speech recognition system software, only the speech recognition controller being new in this alternative.

It will be understood that in this alternative, the special word (“please” in the above example) can be redefined to match alternate “baseform” sequences (as known in WVS—WebSphere™ Voice Server—terminology) which, in this case, are the various truncated endings of the sound. The special word will not be usable in other grammars where it is not permissible to accept truncated utterances.

It will be understood that the above-described preferred embodiment's method of increasing speech recognition performance with speaker phones and car kits provides the following advantages.

While the reliability of recognition may be reduced when compared with a good (microphone unsuppressed or full duplex) telephony situation, especially with short utterances, recognition will be made possible when it is often impossible without this method.

Recognition match scoring will be increased as the low weighting given by deleted initial sounds will be eliminated and therefore confusion of the recognized phrase will be reduced.

The method described with reference to FIG. 3 has the advantage that no modifications are required to the system software but requires the cooperation of the user and may be most appropriate to expert and high value systems where the user has a lot to gain by making the system work for themselves.

The present invention may be realized in hardware, software, or a combination of hardware and software. The present invention may be realized in a centralized fashion in one computer system or in a distributed fashion where different elements are spread across several interconnected computer systems. Any kind of computer system or other apparatus adapted for carrying out the methods described herein is suited. A typical combination of hardware and software may be a general purpose computer system with a computer program that, when being loaded and executed, controls the computer system such that it carries out the methods described herein.

The present invention also may be embedded in a computer program product, which comprises all the features enabling the implementation of the methods described herein, and which when loaded in a computer system is able to carry out these methods. Computer program in the present context means any expression, in any language, code or notation, of a set of instructions intended to cause a system having an information processing capability to perform a particular function either directly or after either or both of the following: a) conversion to another language, code or notation; b) reproduction in a different material form.

This invention may be embodied in other forms without departing from the spirit or essential attributes thereof. Accordingly, reference should be made to the following claims, rather than to the foregoing specification, as indicating the scope of the invention.

Claims

1. A method of speech recognition for use with a system having a microphone, comprising: receiving a signal from the microphone representative of speech to be recognized when the system is in a state where the microphone has been turned off; performing detection of a transition in the signal indicative of switch on of the microphone; and in response to the detection performing speech recognition on the signal with reduced contribution from an initial portion thereof.

2. The method of claim 1, wherein the step of performing speech recognition comprises treating the initial portion as optional speech that is acceptable to have absent from a received utterance within the signal.

3. The method of claim 2, wherein the initial portion comprises at least one of first word and a received and terminal portion of the first word of a spoken series of words contained within said signal.

4. The method of claim 3, further comprising the step of: establishing an utterance initiating word for the system, wherein said first word is said utterance initiating word.

5. The method of claim 1, wherein the step of performing speech recognition comprises performing speech recognition with a predetermined redundant sound.

6. The method of claim 5, wherein said predetermined redundant sound is user configurable.

7. The method of claim 5 further comprising detecting when speech recognition has fallen below a predetermined threshold, and in response thereto requesting a user to speak the predetermined redundant sound.

8. The method of claim 1, wherein the system is a telephone speaker phone system.

9. The method of claim 8, wherein the system is an in-car system.

10. The method of claim 1, further comprising: establishing a configurable parameter for initial word recognition sensitivity, wherein a quantity by which a contribution of the initial portion is reduced is dependent upon a value of said configurable parameter.

11. A speech recognition arrangement for use with a system having a microphone, comprising: means for receiving a signal from the microphone representative of speech to be recognized; means for performing detection of a transition in the signal indicative of switch on of the microphone; and means for performing, in response to the detection, speech recognition on the signal with reduced contribution from an initial portion thereof.

12. The arrangement of claim 11, wherein the means for performing speech recognition comprises means for treating the initial portion as optional speech.

13. The arrangement of claim 11, wherein the means for performing speech recognition comprises performing speech recognition with a predetermined redundant sound.

14. The arrangement of claim 13 further comprising means for detecting when speech recognition has fallen below a predetermined threshold, and in response thereto requesting a user to speak the predetermined redundant sound.

15. The arrangement of claim 11, wherein the system is a telephone speakerphone system.

16. The arrangement of claim 15, wherein the system is an in-car system.

17. A speech recognition method comprising: a speech recognition system receiving a speech input when the speech recognition system is in a state where a microphone input is turned off; identifying an initial portion of the speech input; speech recognizing said speech input in a manner that the identified initial portion has a reduced contribution compared to other portions of the speech input.

18. The speech recognition method of claim 17, further comprising: before said speech recognizing step, discarding said initial portion of the speech input, wherein the initial portion has a reduced contribution because it is not speech recognized.

19. The speech recognition method of claim 17, further comprising: identifying the initial portion as a portion of the speech input proceeding a first detected pause in the speech input.

20. The speech recognition method of claim 17, wherein the initial portion comprises at least one of a first spoken word within the speech input and a received and terminal portion of the first spoken word within the speech input.