A Versatile Recognition Processor Employing HaarLike Feature and Cascaded Classifier

📋 论文概要

该论文提出了一种通用识别处理器，采用Haar-like特征和级联分类器，能够检测和识别图像、视频、声音和加速度信号。在QVGA人脸检测任务中，功耗仅为0.47mW/fps，准确率达81%，比传统处理器功耗低57倍，适用于便携电子和无线传感器网络。

💡 主要创新点

• 提出一种支持多种信号（图像、视频、声音、加速度）的通用识别处理器架构
• 采用Haar-like特征和级联分类器实现低功耗高精度检测，功耗比传统方案降低57倍

🏷 关键词

📄 原文摘要

This paper presents a versatile recognition processor that performs detection and recognition of image, video, sound and acceleration signals, while dissipating 0.15µW/fps to 0.47mW/fps (Fig. 8.2.1). Given the low power dissipation of sub-mW/fps, this processor is suitable for use in portable electronics and wireless sensor networks (WSN) [1]. For instance, it detects human faces from a QVGA image with 81% accuracy and consumes 0.47mW/fps. Power consumption is 57× lower than that of conventional object recognition processors [2, 3] with comparable accuracy (Fig. 8.2.2). A fair comparison, by taking technology differences into account, shows greater than 8× power efficiency. This processor detects speech from very short and low quality sound signals (72ms in 10s, 8kHz, 8b) recorded by a microphone in a sensor node. It also recognizes human activities such as walking, reading and typing from short and low quality 3D acceleration signals (2s in 10s,

👥 作者与机构

Yuya Hanai, Yuichi Hori, Jun Nishimura, Tadahiro Kuroda

Keio University, Yokohama, Japan