Speech recognition method and system
A speech recognition method comprising the steps of: storing multiple recognition models for a vocabulary set, each model distinguished from the other models in response to a Lombard characteristic, detecting at least one speaker utterance in a motor vehicle, selecting one of the multiple recognition models in response to a Lombard characteristic of the at least one speaker utterance, utilizing the selected recognition model to recognize the at least one speaker utterance; and providing a signal in response to the recognition.
1. A speech recognition method, comprising:
storing multiple recognition models for a vocabulary set, each model classified according to a Lombard cluster, wherein each model is distinguishable from the other models based on a Lombard characteristic, the Lombard characteristic representing changes in an utterance pitch and volume in the presence of background noise;
detecting at least one speaker utterance in a motor vehicle;
selecting, via a processing device, one of the multiple recognition models in response to a Lombard characteristic of the at least one speaker utterance;
utilizing the selected recognition model to recognize the at least one speaker utterance; and
providing a signal in response to the recognition.
2. The method according to claim 1 , further comprising building the stored multiple recognition models by:
recording utterances from a plurality of persons, wherein at least some of the recorded utterances are made with background noise audible to each person but not included in the recording of the utterances, and wherein the background noise is varied with different recordings to create a Lombard corpus; and
classifying the recorded utterances in response to the Lombard characteristic to create a plurality of classifications.
3. The method according to claim 2 wherein the building further comprises:
convolving data in the Lombard corpus with a response characteristic of a vehicle;
adding noise to the convolved data; and
training the multiple recognition models based on a result of the adding, wherein each recognition model corresponds to one of the plurality of classifications.
4. The method of claim 3 wherein the noise added to the convolved data represents vehicle ambient noise at various operating conditions of the vehicle.
5. The method of claim 2 wherein the Lombard characteristic includes a Lombard level of the background noise.
6. The method of claim 1 wherein the Lombard characteristic includes a curve representing speaker utterances correlated to changing background noise.
7. The method of claim 6 , further comprising storing the detected speaker utterances in a memory by determining a Lombard curve of a plurality of detected speaker utterances, wherein the selecting of the one of the multiple recognition models is in response to the determined Lombard curve.
8. The method of claim 1 wherein the Lombard characteristic includes a partial curve of speaker utterances in response to changing background noise.
9. The method of claim 1 wherein the Lombard characteristic includes a background noise level audible to a speaker during an utterance.
10. The method of claim 1 , further comprising selecting a default model if a model matching the Lombard characteristic of the at least one speaker utterance is not available.
11. The method of claim 1 , further comprising transmitting data responsive to the at least one speaker utterance to a remote station, wherein the selecting of the one of the multiple recognition models is performed at the remote station.
12. The method of claim 11 , further comprising downloading the selected recognition model to an in-vehicle device.
13. The method of claim 11 wherein the recognizing is performed at the remote station.
14. The method of claim 1 wherein the recognizing is performed by an in-vehicle device.
15. A speech recognition system, comprising:
memory containing multiple recognition models for a vocabulary set, each recognition model trained according to a Lombard cluster of utterances, wherein each model is distinguishable from the other models based on a Lombard characteristic, the Lombard characteristic representing changes in an utterance pitch and volume in the presence of background noise;
a sound-detecting device receiving at least one speaker utterance; and
a processing device containing control structure executed to: select one of the multiple recognition models in response to a Lombard characteristic of the at least one speaker utterance; utilize the selected recognition model to recognize the at least one speaker utterance; and provide a signal in response to the recognizing.
16. The system of claim 15 wherein the sound-detecting device is in a motor vehicle.
17. The system of claim 16 wherein the processing device is located at a station remote from the motor vehicle.
18. The system of claim 15 wherein the processing device is integrated into a motor vehicle.
19. A speech recognition system, comprising:
a vocabulary recording subsystem for recording utterances of a desired vocabulary from a plurality of speakers and storing data from the recorded utterances as a corpus, wherein background noise is audible to each speaker but not contained in the recorded utterances;
a mixing device for mixing various background sounds with the recorded utterances; and
a data structure containing at least two models of the desired vocabulary, each of the at least two models classified according to a Lombard cluster, wherein each of the models are responsive to the mixing device, and wherein each of the models is distinguishable from the other based on a Lombard characteristic of at least a portion of the corpus, the Lombard characteristic representing changes in an utterance pitch and volume in the presence of background noise.
20. The speech recognition system of claim 19 wherein at least one copy of the data structure is located in a motor vehicle.
21. The speech recognition system of claim 19 , further comprising a convolving device configured to convolve the recorded utterances with a signal representative of a vehicle acoustic response and to provide the recorded utterances to the mixing device.