Microphone Array Beamforming for Reliable Voice Wake-Up in Noise

The patent segments the beamforming process into multiple sequential stages: initial broad beam formation for orientation screening, followed by refined beamforming based on screened orientation. This segmentation allows the system to first identify potential orientations without requiring perfect denoising, then apply targeted denoising in subsequent stages, resolving the contradiction between operating without positioning hardware and achieving reliable denoising.

The patent performs preliminary orientation screening using broad beamforming before applying refined denoising algorithms. This preliminary action identifies candidate orientations that guide subsequent targeted denoising efforts, allowing the system to achieve reliable voice awakening without requiring additional positioning hardware by preparing the processing pipeline in advance with orientation hypotheses.

The patent enables the microphone array system to perform its own orientation determination and beamforming optimization using only the audio signals it receives. The system self-determines orientations through broad beamforming and screening without requiring external positioning devices, then uses this self-acquired orientation information to guide refined beamforming, eliminating the need for additional positioning hardware while achieving effective denoising.

The patent makes the microphone array system multi-functional by having it perform both initial orientation screening and refined denoising operations using the same hardware components. The beamforming system serves multiple purposes: initial broad beam formation for orientation identification, subsequent refined beamforming for denoising, and final voice awakening processing, eliminating the need for separate positioning hardware.

The patent implements dynamic beamforming where the beam orientation and width are adjusted based on the screening results and signal characteristics. The system dynamically transitions from broad beamforming for orientation screening to refined beamforming for denoising, adapting its processing strategy to the specific acoustic environment and improving versatility across different scenarios without requiring additional hardware.

The patent changes key beamforming parameters dynamically: initial broad beam width for orientation screening, then refined beam width and orientation based on screening results. This parameter adaptation allows the system to handle diverse acoustic scenarios effectively, improving scenario coverage and adaptability while maintaining simple hardware architecture.