← 返回 JSSC 论文列表
📄 下载 JSSC 原文 PDF
JSSC 2013第8期Other32nm

A 23 nJFrame V oice Activity Detector-Based Audio Front-End for Context-Aware Sy

一款基于语音活动检测的低功耗音频前端芯片,适用于嵌入式SoC。
23nJ/帧,50uW功耗,32nm CMOS工艺
语音活动检测低功耗设计音频前端嵌入式SoC近阈值电压
创新点1:512点FFT硬件加速 - 采用专用硬件加速器实现512点FFT运算,显著提升音频信号处理的实时性,相比软件实现能效提升9.4倍,特别适合嵌入式系统的低功耗要求(方法创新)
创新点2:可编程滤波器设计 - 通过灵活配置的滤波器组实现多频段音频处理,支持动态调整截止频率和增益,适应不同环境噪声条件下的语音检测需求(电路创新)
创新点3:近阈值电压(NTV)技术 - 在32nm CMOS工艺下实现双电压域设计,核心数据路径工作于近阈值电压区域,功耗低于50μW,达到23nJ/帧的超低能耗指标(系统级能效创新)
创新点4:集成化噪声基底估计器 - 硬件集成的动态噪声评估模块,通过实时环境噪声分析自动调整检测阈值,提升语音活动检测的鲁棒性(算法-硬件协同创新)
Abstract
Advanced human-machine interfaces require im- proved embedded sensors that can seamlessly interact with the user. V oice-based communication has emerged as a promising interface for next generation mo bile, automotive and hands-free devices. Presented here is such an audio front-end with Voice Activity Detection (V AD) hardware targeted for low-power em- bedded SoCs, featuring a 512 pt FFT, programmable filters, noise floor estimator and a decision engine which has been fabricated in 32 nm CMOS. T