Speech recognition for vehicle voice commands

1. A vehicle comprising:

a microphone to collect a signal including a voice command;

a memory; and

a controller configured to:

determine an initial identification of the voice command by feeding the signal into a first automatic speech recognition (ASR) engine, wherein the initial identification of the voice command is a final output of the first ASR engine, wherein to determine the initial identification comprises identifying, via an acoustic model of the first ASR engine, one or more phonemes of a dialect within the signal;

determine habits by feeding user history into a habits engine;

identify, based on determining that a confidence level associated with the initial identification is greater than a confidence threshold and a noise level associated with the initial identification is greater than a noise threshold, the voice command by feeding the signal, the initial identification, and the habits into a second ASR engine, wherein to identify the voice command comprises identifying, via a language model of the second ASR engine, one or more words within the signal by determining word probability distributions based on the one or more phonemes identified by the acoustic model; and

perform a vehicle function based on the voice command.

2. The vehicle of claim 1 , wherein the controller utilizes the second ASR engine to identify the voice command responsive to determining that the initial identification corresponds with the confidence level that is less than the confidence threshold.

3. The vehicle of claim 1 , wherein the controller identifies the initial identification as the voice command responsive to determining that:

the initial identification corresponds with the confidence level that is greater than the confidence threshold; and

the noise level of the signal is less than the noise threshold.

4. The vehicle of claim 1 , wherein the first ASR engine further includes:

the language model.

5. The vehicle of claim 1 , wherein the language model includes a deep neural network.

6. The vehicle of claim 1 , wherein the habits engine includes a pattern recognition algorithm.

7. The vehicle of claim 1 , further including one or more input devices, wherein the controller determines the user history based on user inputs received by the one or more input devices.

8. A system comprising:

a vehicle to operate based on a voice command, the vehicle includes a microphone to collect a signal including the voice command; and

a remote server, in communication with the vehicle, to:

determine an initial identification of the voice command via a first engine based on the signal, wherein the initial identification of the voice command is a final output of the first engine, wherein to determine the initial identification comprises identifying, via an acoustic model of the first engine, one or more phonemes of a dialect within the signal;

determine habits via a habits engine; and

identify, based on determining that a confidence level associated with the initial identification is greater than a confidence threshold and a noise level associated with the initial identification is greater than a noise threshold, the voice command for the vehicle via a second engine based on the signal, the initial identification, and the habits, wherein the first engine is different from the second engine, wherein to identify the voice command comprises identifying, via a language model of the second engine, one or more words within the signal by determining word probability distributions based on the one or more phonemes identified by the acoustic model.

9. The system of claim 8 , wherein the remote server utilizes the second engine to identify the voice command responsive to determining that the initial identification corresponds with the confidence level that is less than the confidence threshold.

10. The system of claim 8 , wherein the remote server identifies the initial identification as the voice command responsive to determining that:

the initial identification corresponds with the confidence level that is greater than the confidence threshold; and

the noise level of the signal is less than the noise threshold.

11. The system of claim 8 , wherein the first engine further includes:

the language model.

12. The system of claim 8 , wherein the language model includes a deep neural network.

13. The system of claim 8 , wherein the habits engine includes a pattern recognition algorithm.

14. The system of claim 8 , wherein the vehicle includes one or more input devices, wherein the remote server determines user history to be fed into the habits engine based on user inputs received by the one or more input devices.

15. A method comprising:

collecting, via a vehicle microphone, a signal that includes a voice command;

determining an initial identification of the voice command by feeding the signal into a first automatic speech recognition (ASR) engine, wherein the initial identification of the voice command is a final output of the first ASR engine, wherein determining the initial identification comprises identifying, via an acoustic model of the first ASR engine, one or more phonemes of a dialect within the signal;

determining habits by feeding user history into a habits engine;

identifying, based on determining that a confidence level associated with the initial identification is greater than a confidence threshold and a noise level associated with the initial identification is greater than a noise threshold, the voice command by feeding the signal, the initial identification, and the habits into a second ASR engine, wherein the first ASR engine is different from the second ASR engine, wherein identifying the voice command comprises identifying, via a language model of the second ASR engine, one or more words within the signal by determining word probability distributions based on the one or more phonemes identified by the acoustic model; and

performing, via a processor, a vehicle function based on the voice command.

16. The method of claim 15 , wherein the voice command is identified via the second ASR engine responsive to determining that the initial identification corresponds with the confidence level that is less than the confidence threshold.

17. The method of claim 15 , further including identifying the initial identification as the voice command responsive to determining that:

the confidence level that is greater than the confidence threshold; and

the noise level of the signal is less than the noise threshold.

18. The method of claim 15 , wherein the language model includes a deep neural network.

19. The method of claim 15 , wherein the habits engine includes a pattern recognition algorithm.