A Target-Separable BWN Inspired Speech Recognition Processor with Low-power Precision-adaptive Approximate Computing
Bo Liu1,b, Hao Cai1,a, Xuan Zhang1,c, Haige Wu1,d, Anfeng Xue1,e, Zilong Zhang1,f, Zhen Wang2 and Jun Yang1,g
1National ASIC System Engineering Research Center, Southeast University, Nanjing, China
ahao.cai@seu.edu.cn
bliubo_cnasic@seu.edu.cn
cz_xuan@seu.edu.cn
dwu_haige@seu.edu.cn
exue_anfeng@seu.edu.cn
fzhangzilong@seu.edu.cn
gdragon@seu.edu.cn
2Nanjing Prochip Electronic Technology Co. Ltd. Nanjing, China
fanstics@prochip.com.cn
ABSTRACT
This paper proposes a speech recognition processor based on a target-separable binarized weight network (BWN), capable of performing both speaker verification (SV) and keyword spotting (KWS). In traditional speech recognition system, the SV based on traditional model and the KWS based on neural networks (NN) model are two independent hardware modules. In this work, both SV and KWS are processed by the proposed BWN with unified training and optimization framework which can be performed for various application scenarios. By the systemarchitecture co-design, SV and KWS share most of the network parameters, and the classification part is calculated separately according to different targets. An energy-efficient NN accelerator which can be dynamically reconfigured to process different layers of the BWN with splitting calculation of frequency domain convolution is proposed. SV and KWS can be achieved with only one time calculation of each input speech frame, which greatly improves the computing energy efficiency. The computing units of the NN accelerator are optimized using precision-adaptive approximate computing method with Dual-VDD to further reduce the energy cost. Compared to state-of-the-arts, this work can achieve about 4× reduction in power consumption while maintaining high system adaptability and accuracy.
Keywords: Speaker verification, keyword spotting, precisionadaptive approximate computing, binarized weight network, frequency domain convolution, target-separable.
