Driver interface with voice and image control
A driver interface for use within an automobile provides responses to voice commands issued for example by a driver of the automobile. The interface includes a camera and microphone for capturing image data such as gestures and audio data from the automobile driver. The image data and audio data are processed to extract image and linguistic features from the image and audio data, which image and linguistic features are processed to interpret and infer a meaning of the voice command.
1 . A driver interface system comprising:
a camera interface enabled to capture images of a driver and an environment surrounding the driver;
an image processor enabled to extract, from captured images of the driver and an environment surrounding the driver, visual features, the visual features being an abstraction of the captured images, the abstraction being a compressed representation of the captured images;
a microphone interface enabled to capture speech;
a speech recognizer enabled to extract, from the captured speech, linguistic features;
a natural language processor enabled to infer a voice command from the linguistic features and the visual features; and
an output interface for rendering a response to the voice command; and
wherein the speech recognizer uses a machine-learned linguistic model based on a combination of audio feature tensors and visual feature tensors.
2 . The driver interface system of claim 1 wherein inferring the voice command from the linguistic features and the visual features is configured by a set of acoustic model configurations parameters.
3 . The driver interface system of claim 1 further comprising a data interface enabled to transmit visual features and captured speech to a server, wherein the speech recognizer extracts linguistic features by making an application programming interface (API) call to the server through the data interface.
4 . The driver interface system of claim 3 wherein the data interface is a wireless network interface.
5 . The driver interface system of claim 1 further comprising a data interface enabled to transmit visual features and captured speech to a server, wherein the natural language processor infers the voice command by making an application programming interface (API) call to the server through the data interface.
6 . The driver interface system of claim 5 wherein the data interface is a wireless network interface.
7 . The driver interface system of claim 1 wherein the visual features indicate a gesture.
8 . The driver interface system of claim 1 wherein the response is rendered by projection onto a display screen.
9 . A driver interface system comprising:
a camera interface enabled to capture images of a driver and an environment surrounding the driver;
a neural network image processor enabled to extract, from captured images of the driver and an environment surrounding the driver, visual features, the visual features being an abstraction of the captured images determined by the neural network image processor, the abstraction being a compressed representation of the captured images with reduced information content as compared to the captured images;
a microphone interface enabled to capture speech;
a speech recognizer enabled to extract, from the captured speech, linguistic features; and
a natural language processor enabled to infer a voice command from the linguistic features and the visual features, including visual features absracted from an environment surrounding the driver;
wherein the image processor, the speech recognizer, and the natural language processor are jointly configured by joint training, such that the extraction of the visual feature tensors and audio feature tensors is optimized in combination with the natural language processor to minimize errors in inferring the voice command.
10 . The driver interface system of claim 9 wherein inferring the voice command from the linguistic features and the visual features is configured by a set of acoustic model configurations parameters.
11 . The driver interface system of claim 9 further comprising a data interface enabled to transmit visual features and captured speech to a server, wherein the speech recognizer extracts linguistic features by making an application programming interface (API) call to the server through the data interface.
12 . The driver interface system of claim 11 wherein the data interface is a wireless network interface.
13 . The driver interface system of claim 9 further comprising a data interface enabled to transmit visual features and captured speech to a server, wherein the natural language processor infers the voice command by making an application programming interface (API) call to the server through the data interface.
14 . The driver interface system of claim 13 wherein the data interface is a wireless network interface.
15 . The driver interface system of claim 9 wherein the speech recognizer uses a machine-learned linguistic model based on a combination of audio feature tensors and visual feature tensors.
16 . The driver interface system of claim 9 wherein the visual features indicate a gesture.
17 . The driver interface system of claim 9 further comprising an output interface for rendering a response to the voice command.
18 . The driver interface system of claim 17 wherein the response is rendered by projection onto a display screen.
19 . A driver interface system comprising:
a camera interface enabled to capture images of a driver and an environment surrounding the driver;
an image processor enabled to extract, from captured images of the driver and an environment surrounding the driver, visual feature tensors, the visual feature tensors being an abstraction of the captured images, the abstraction being a compressed representation of the captured images with reduced information content as compared to the captured images;
a microphone interface enabled to capture speech;
a speech recognizer enabled to extract, from the captured speech, audio feature tensors;
a natural language processor enabled to concatonate or merge the visual feature vectors and audio feature vectors to infer a voice command;
wherein the image processor, the speech recognizer, and the natural language processor are jointly configured by joint training, such that the extraction of the visual feature tensors and audio feature tensors is optimized in combination with the natural language processor to minimize errors in inferring the voice command; and
an output interface for rendering a response to the voice command.