A Multimode 157μW 4-Channel 80dBA-SNDR Speech-Recognition Frontend With Self-DOA Correction Adaptive Beamformer

Abstract

Beamforming with multiple microphones is essential for Automatic Speech Recognition (ASR) in earbuds, cell phones, and smart speakers. Although fixed delay-and-sum (DAS) beamforming is simple to implement, it only suppresses noise from a fixed direction of arrival (DoA) [1]; hence, it is ineffective in real varying noise conditions. Reference [2] implements ultra-low-power keyword spotting (KWS) with noise suppression, but the lack of an ADC and beamforming limit practical application. On the other hand, adaptive beamforming (ABF) actively adjusts nulls to suppress varying noise sources. Adaptive beamforming with a trained DNN is promising [3] but requires extensive training data and high power consumption and is not applicable for battery-operated systems. Conventional adaptive beamforming [4 – 5] (Fig. 32.5.1) adaptively reduces noise and interference in the output of a fixed DAS beamformer. Although conventional ABF is effective and compact, it is hampered by: 1) high DSP power consumption due to high ADC sampling rate and the need for complex calculations, especially in the blocking matrix (BM); 2) target signal direction errors in DAS cause severe signal distortion; and 3) worst-case input-SNR design causes high ADC and DSP power regardless of actual signal conditions.