Speech recognition for premature enunciation
View Patent ↗Methods of automatic speech recognition for premature enunciation. In one method, a) a user is prompted to input speech, then b) a listening period is initiated to monitor audio via a microphone, such that there is no pause between the end of step a) and the beginning of step b), and then the begin-speaking audible indicator is communicated to the user during the listening period. In another method, a) at least one audio file is played including both a prompt for a user to input speech and a begin-speaking audible indicator to the user, b) a microphone is activated to monitor audio, after playing the prompt but before playing the begin-speaking audible indicator in step a), and c) speech is received from the user via the microphone.
1. A method of automatic speech recognition, comprising the steps of:
a) prompting a user to input speech;
b) after step a), initiating a listening period to monitor audio via a microphone, such that there is no pause between the end of step a) and the beginning of step b); and then
c) communicating a begin-speaking audible indicator to the user during the listening period;
d) receiving, via the microphone, signals corresponding to input speech from the user and signals corresponding to the begin-speaking audible indicator; and
e) determining if the input speech was prematurely enunciated;
f) applying a digital filter to filter out signals corresponding to the begin-speaking audible indicator, if it is determined in step e) that the user's input speech was prematurely enunciated;
g) estimating speech energies inadvertently filtered from the speech; and
h) appending the identified speech energies to the received speech signals.
2. The method of claim 1 , wherein step g) includes applying context-dependent phoneme models to estimate the speech energies.
3. The method of claim 2 , wherein the context-dependent phoneme models include maximum mutual information models.
4. The method of claim 2 , further comprising the step of i) decoding the received input speech signals with the speech energies appended in step h) to produce a plurality of hypotheses for the user's input speech.
5. The method of claim 4 , further comprising the step of j) post-processing the plurality of hypotheses to identify one of the hypotheses as the user's input speech.
6. The method of claim 1 , further comprising the step of decoding the received input speech signals to produce a plurality of hypotheses for the user's input speech, if it is determined in step e) that the user's input speech was not prematurely enunciated.
7. The method of claim 6 , further comprising the step of post-processing the plurality of hypotheses to identify one of the hypotheses as the user's input speech.
8. A method of automatic speech recognition, comprising the steps of:
a) determining whether a prompt for a user to input speech is concatenated from at least two audio files;
b) based on the determination of step (a), selecting at least two audio files to be played, wherein one of the audio files includes both a prompt for a user to input speech and a begin-speaking audible indicator to the user and at least one other audio file includes another prompt for the user to input speech and excludes the begin-speaking audible indicator;
c) activating a microphone to monitor audio, after playing the prompts but before playing the begin-speaking audible indicator in step a); and
d) receiving speech from the user via the microphone.
9. The method of claim 8 , further comprising the step of:
d) decoding the speech received from the user using at least one speech recognition model that includes the begin-speaking audible indicator.
10. The method of claim 8 , wherein step a) includes playing the audio file including the begin-speaking audible indicator after the audio file excluding the begin-speaking audible indicator.
11. The method of claim 8 , wherein step a) includes playing a single audio file including the begin-speaking audible indicator.