CONTINUOUS VOICE SENSING

Provided are methods and systems for continuous voice sensing. An example method allows for detecting and buffering, by a first module, a key phrase in an acoustic signal. Responsive to the detection, the method includes sending an interrupt to a second module and switching the first module to an omnidirectional microphone mode. Upon receiving the interrupt, the second module is operable to boot up from a low power mode to an operational mode. While the second module is booting up, the first module is operable to continue to buffer a clean speech output generated from an acoustic signal captured by at least one omnidirectional microphone. After the second module is booted, an indication may be sent to the first module that the second module is ready to exchange data through a fast connection. Upon receiving the indication, the buffered clean speech output may be sent to the second module.