To invoke a virtual assistant, a user needs to utter a pre-configured hotword. In devices that provide virtual assistant software, such utterance is detected by dedicated hardware, e.g., a low power digital signal processor (DSP). However, speaking a hotword to trigger a virtual assistant is sometimes unnatural, awkward, or inconvenient. This disclosure describes a two-stage technique that enables virtual assistant invocation without the need for the user to speak a hotword for each query. With user permission, data from multiple sensors that are integrated within a user device and/or connected to the device via fast communication protocols are utilized to detect user intent to issue a command to the virtual assistant.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.