Computerized device with voice command input capability
A computerized device with voice command capability processed remotely includes a low power processor, executing a loose algorithmic model to recognize a wake word prefix in a voice command, the loose model having a low false rejection rate but suffering a high false acceptance rate, and a second processor which can operate in at least a low power/low clock rate mode and a high power/high clock rate mode. When the first processor determines the presence of the wake word, it causes the second processor to switch to the high power/high clock rate mode and to execute a tight algorithmic model to verify the presence of the wake word. By using the two processors in this manner, the average overall power required by the computerized device is reduced, as is the amount of waste heat generated by the system.
1. A method of recognizing a wake word for voice commands to a computerized device, comprising the steps of:
(i) receiving at the computerized device an audio signal from at least one microphone;
(ii) processing the received audio signal with a first processor in the computerized device, the first processor placing a copy of the processed received audio signal into a circular buffer of a preselected size and the first processor executing a first voice recognition algorithmic model to detect the presence of a predefined wake word, the first voice recognition algorithmic model selected to provide a predefined relative low level of false non-matches of the predefined wake word at the cost of a higher than predefined level of false matches of the predefined wake word;
(iii) upon the first processor determining a match of the predefined wake word, the first processor providing a second signal to a second processor in the computerized device, the second processor normally operating at a first rate having a first computational capacity, the second signal causing the second processor to commence operating at a second rate having a second computational capacity greater than the first computational capacity, and the second processor:
(a) copying the contents of the circular buffer into a second buffer;
(b) copying the processed received audio signal into a third buffer;
(c) executing a second voice recognition algorithmic model on the contents of the second buffer to verify the presence of the predefined wake word, the second voice recognition algorithmic model requiring greater computational processing than the first voice recognition algorithmic model to achieve a predefined relatively low level of both false non-matches and false matches of the predefined wake word; and
(d) upon completion of analyzing the contents of the second buffer with the second voice recognition algorithmic model, if the second voice recognition algorithmic model determines that the predefined wake word is not present in the second buffer, returning the second processor to operate at the first rate and, if the second voice recognition algorithmic model determines that the predefined wake word is present in the second buffer, then forwarding the contents of the second buffer and the third buffer to a voice processing service located remote from the computerized device, the voice processing service operable to receive and process voice commands.
2. The method of claim 1 further comprising the step of the second processor activating an activity indicator when the second processor determines, at (c) that the predefined wake word is present in the second buffer.
3. The method of claim 1 where the voice processing service executes a third voice recognition algorithmic model requiring greater computational processing than the second voice recognition algorithmic model, the voice processing service executing the third voice recognition algorithmic model on the copy of the second buffer received at the voice processing service to verify the presence of the wake word therein and, if the third voice recognition algorithmic model does not verify the presence of the wake word, then the voice processing service sending a message to the computerized device indicating that the wake word was not present and the second processor returning to operating at the first rate and, if the third voice recognition algorithmic model does verify the presence of the wake word, then the voice processing service processing the contents of the third buffer.
4. A computerized device comprising:
at least one microphone to capture user voices;
a first processor to digitize and process audio received from the at least one microphone and to store a copy of the processed audio in a circular buffer and to execute a first voice recognition algorithmic model to detect the presence of a predefined wake word in the circular buffer, the first voice recognition algorithmic model selected to provide a predefined relative low level of false non-matches of the predefined wake word at the cost of a higher than predefined level of false matches of the predefined wake word;
a second processor normally operating at a first rate having a first computational capacity and responsive to a signal from the first processor indicating that the first voice recognition algorithmic model has detected the presence of the wake word in the circular buffer such that the second processor commences operation at a second rate having a greater computational capacity than the capacity at the first rate, the second processor receiving a copy of the contents of the circular buffer from the first processor and receiving and buffering a copy of the processed received audio stream in a second buffer, the second processor executing a second voice recognition algorithmic model on the copy of the contents of the circular buffer to verify the presence of the predefined wake word, the second voice recognition algorithmic model requiring greater computational processing than the first voice recognition algorithmic model and being selected to achieve a predefined relatively lower level of both false non-matches and false matches of the predefined wake word than achieved by the first processor; and
a data communications module operable to provide data communication between the computerized device and a remote voice processing service, the data communications providing the voice processing service with the copy of the contents of the circular buffer and the contents of second buffer to the voice processing service when the second voice recognition algorithmic model verifies the presence of the wake word in the copy of the contents of the circular buffer.