Acoustic and domain based speech recognition for vehicles
A processor of a vehicle speech recognition system recognizes speech via domain-specific language and acoustic models. The processor further, in response to the acoustic model having a confidence score for recognized speech falling within a predetermined range defined relative to a confidence score for the domain-specific language model, recognizes speech via the acoustic model only.
1. A speech recognition method comprising:
executing, by a processor, a vehicle command identified from a signal containing speech according to a recognition hypothesis selected from a plurality of hypotheses that are each based on a product of a common speech domain pair, the common speech domain pair including (i) one of a plurality of domain-specific language model confidence scores derived from the speech and application of a machine-learning algorithm to vehicle state inputs and (ii) one of a plurality of acoustic model confidence scores; and
proportionally decreasing, by the processor, the one of the plurality of domain-specific language model confidence scores prior to generation of the product based on the one of the plurality of acoustic model confidence scores falling within a predetermined range that is defined relative to the one of the plurality of domain-specific language model confidence scores.
2. The method of claim 1 wherein the machine-learning algorithm is an artificial neural network.
3. The method of claim 2 wherein the artificial neural network has an output that is related to the commands.
4. The method of claim 1 , wherein the vehicle state inputs include weather or traffic.
5. The method of claim 1 , wherein the vehicle state inputs include nomadic devices in proximity to the vehicle.
6. The method of claim 1 , wherein the vehicle state inputs include conversational history.
7. A speech recognition system comprising:
a processing device programmed to
execute a vehicle command identified from a signal containing speech according to a recognition hypothesis selected from a plurality of hypotheses that are each based on a product of a common speech domain pair, the common speech domain pair including (i) one of a plurality of domain-specific language model confidence scores derived from the speech and application of a machine-learning algorithm to vehicle state inputs and (ii) one of a plurality of acoustic model confidence scores, and
proportionally decreasing the one of the plurality of domain-specific language model confidence scores prior to generation of the product based on the one of the plurality of acoustic model confidence scores falling within a predetermined range that is defined relative to the one of the plurality of domain-specific language model confidence scores.
8. The system of claim 7 wherein the machine-learning algorithm is an artificial neural network.
9. The system of claim 8 wherein the artificial neural network has an output that is related to the commands.
10. The system of claim 7 , wherein the vehicle state inputs include weather or traffic.
11. The system of claim 7 , wherein the vehicle state inputs include nomadic devices in proximity to the vehicle.
12. The system of claim 7 , wherein the vehicle state inputs include conversational history.