Techniques for spatially selective wake-up word recognition and related systems and methods

A system for detecting a designated wake-up word in a physical space having a plurality of acoustic zones, the system comprising:

• at least one processor configured to:
• receive a first acoustic signal comprising acoustic information from a plurality of microphones comprising a first microphone in a first acoustic zone of the plurality of acoustic zones and a second microphone in a second acoustic zone of the plurality of acoustic zones;
• process the first acoustic signal to identify content of the first acoustic signal originating from each of the plurality of acoustic zones;
• generate, based on the first acoustic signal, a plurality of second acoustic signals, each of the plurality of second acoustic signals corresponding to one of the first acoustic zone and the second acoustic zone and comprising content corresponding to the content identified as originating from the one acoustic zone, wherein the number of the second acoustic signals is less than the number of the acoustic zones;
• separately perform automatic speech recognition on one or more of the plurality of second acoustic signals to determine whether the designated wake-up word was spoken;
• select, in response to the designated wake-up word being spoken, the first acoustic zone of the plurality of acoustic zones based on, for each of one or more of the plurality of second acoustic signals, whether the designated wake-up word was spoken in the first acoustic zone;
• perform additional automatic speech recognition of sound originating from the first acoustic zone to target the first acoustic zone for subsequent automatic speech recognition;
• provide a plurality of third acoustic signals corresponding to the content identified as originating from a respective one of the plurality of acoustic zones, wherein a number of third acoustic signals is equal to the number of acoustic zones;
• and
• select the plurality of second acoustic signals as a subset of the plurality of third acoustic signals based at least in part upon data from at least one hardware sensor, wherein the hardware sensor is an occupant detection sensor to detect whether an occupant is located within one of the plurality of acoustic zones.

A method for detecting a designated wake-up word in a physical space having a plurality of acoustic zones, the method comprising:

• receive a first acoustic signal comprising acoustic information from a plurality of microphones comprising a first microphone in a first acoustic zone of the plurality of acoustic zones and a second microphone in a second acoustic zone of the plurality of acoustic zones;
• processing the first acoustic signal to identify content of the first acoustic signal originating from each of the plurality of acoustic zones;
• generating, based on the first acoustic signal, a plurality of second acoustic signals, each of the plurality of second acoustic signals corresponding to one of the first and second acoustic zones and comprising content corresponding to the content identified as originating from the one acoustic zone;
• separately performing automatic speech recognition on one or more of the plurality of second acoustic signals including using a different acoustic model for each of the plurality of second acoustic signals to determine whether the designated wake-up word was spoken, wherein the acoustic model for each of the plurality of second acoustic signals includes a speech recognition model for the acoustic environment of an associated acoustic zone;
• identify one or more spoken commands within the sound originating from the first acoustic zone;
• and
• determine whether that the first acoustic zone is authorized to perform the one or more spoken commands.