Motion adaptive speech recognition for enhanced voice destination entry
A method or associated system for motion adaptive speech processing includes dynamically estimating a motion profile that is representative of a user's motion based on data from one or more resources, such as sensors and non-speech resources, associated with the user. The method includes effecting processing of a speech signal received from the user, for example, while the user is in motion, the processing taking into account the estimated motion profile to produce an interpretation of the speech signal. Dynamically estimating the motion profile can include computing a motion weight vector using the data from the one or more resources associated with the user, and can further include interpolating a plurality of models using the motion weight vector to generate a motion adaptive model. The motion adaptive model can be used to enhance voice destination entry for the user and re-used for other users who do not provide motion profiles.
1. A method for motion adaptive speech processing for voice destination entry, the method comprising:
dynamically estimating a motion profile that is representative of a user's motion based on data from one or more resources associated with the user, the motion profile being a collection of snap shots of the user's position and motion, the data from the one or more resources including sensor data and data from a non-speech resource associated with the user;
wherein dynamically estimating the motion profile includes computing a motion weight vector using the data from the one or more resources associated with the user and interpolating a plurality of models using the motion weight vector to generate a motion adaptive model, wherein at least one of the models is associated with a geographic area, and wherein interpolating the models results in a probability that the user is or will be located in the geographic area; and
effecting processing of a speech signal received from the user, the processing taking into account the estimated motion profile by constraining a search space to user-relevant destinations based on the motion adaptive model to produce an interpretation of the speech signal.
2. The method according to claim 1 , wherein the sensor data include at least one member selected from the group consisting of position, speed, acceleration, direction, and a combination thereof.
3. The method according to claim 2 , wherein the data from the non-speech resource includes at least one member selected from the group consisting of navigation system data, address book data, calendar data, motion history data, crowd sourced data, configuration data, and a combination thereof.
4. The method according to claim 1 , wherein computing the motion weight vector includes determining a relation between the non-speech resource and a language resource associated with the plurality of models.
5. The method according to claim 1 , wherein dynamically estimating the motion profile further includes interpolating the motion adaptive model with a background model.
6. The method according to claim 1 , wherein the speech signal includes at least one of a voice audio signal, a video signal and data from gestures or text entry.
7. A system for motion adaptive speech processing for voice destination entry, the system comprising:
a motion profile estimator at a client, the estimator configured to estimate a motion profile that is representative of a user's motion dynamically based on data from one or more resources associated with the user, the motion profile being a collection of snap shots of the user's position and motion, the data from the one or more resources including sensor data and data from a non-speech resource associated with the user;
wherein the motion profile estimator is configured to compute a motion weight vector using the data from the one or more resources associated with the user and to interpolate a plurality of models using the motion weight vector to generate a motion adaptive model, wherein at least one of the models is associated with a geographic area, and wherein the motion profile estimator interpolates the models to produce a probability that the user is or will be located in the geographic area; and
a processor configured to effect processing of a speech signal received from the user, the processing taking into account the estimated motion profile by constraining a search space to user-relevant destinations based on the motion adaptive model to produce an interpretation of the speech signal.
8. The system according to claim 7 , wherein the motion profile estimator is further configured to interpolate the motion adaptive model with a background model.
9. The system according to claim 7 , wherein the processor is configured to perform automatic speech recognition (ASR) of the speech signal at the client using the estimated motion profile, the ASR producing the interpretation of the speech signal.
10. The system according to claim 7 , wherein the processor is configured to send the speech signal and estimated motion profile to a cloud service to perform automatic speech recognition (ASR) of the speech signal using the estimated motion profile, the ASR producing the interpretation of the speech signal.
11. The system according to claim 7 , wherein the processor is configured to send the speech signal to a cloud service for automatic speech recognition (ASR), receive results of the ASR from the cloud service, and re-rank the results using the estimated motion profile to produce the interpretation of the speech signal.
12. A computer program product comprising a non-transitory computer readable medium storing instructions for performing a method for motion adaptive speech processing for voice destination entry, the instructions, when executed by a processor, cause the processor to:
dynamically estimate a motion profile that is representative of a user's motion based on data from one or more resources associated with the user, the motion profile being a collection of snap shots of the user's position and motion, the data from the one or more resources including sensor data and data from a non-speech resource associated with the user;
wherein dynamically estimating the motion profile includes computing a motion weight vector using the data from the one or more resources associated with the user and interpolating a plurality of models using the motion weight vector to generate a motion adaptive model, wherein at least one of the models is associated with a geographic area, and wherein interpolating the models results in a probability that the user is or will be located in the geographic area; and
effect processing of a speech signal received from the user, the processing taking into account the estimated motion profile by constraining a search space to user-relevant destinations based on the motion adaptive model, to produce an interpretation of the speech signal.