基于数字信号处理器的婴儿声音识别系统的设计与实现  

Design and implementation of system to recognize babies' crying based on digital signal processor systems

作　　者：顾国良 许鹏[1] 沈晓燕[1] GU Guoliang;XU Peng;SHEN Xiaoyan(School of Electronics and Information,Nantong University,Nantong 226019,China)

机构地区：[1]南通大学电子信息学院,江苏南通226019

出　　处：《生物医学工程研究》2018年第3期276-280,共5页Journal Of Biomedical Engineering Research

基　　金：国家自然科学基金资助项目(81371663;61534003);江苏省研究生创新训练项目(KYCX17-1917);江苏省高校品牌专业建设工程资助项目(PPZY2015B135);南通市科技计划项目(MS12016026)

摘　　要：为提高对婴儿健康监护的质量,本研究提出在育婴箱中增加婴儿哭声识别功能,该系统主要以TI公司的数字信号处理器(DSP)芯片TMS320DM643和多媒体音频编解码芯片TLC320AIC23B为硬件核心,设计一套实时的婴儿哭声识别系统。本系统使用拾音器采集婴儿哭声,音频解码芯片对声音信号进行处理,然后发送给DSP芯片,DSP芯片对声音信号进行预处理,然后用优化的自相关函数算法提取特征参数-线性预测系数(LPC),使用动态时间规整(DTW)识别算法,实现对婴儿的哭声准确识别,并经串口向上位机发送识别结果。该系统在实际测试后表明,婴儿啼哭状态识别准确率可达97.1%,在婴儿护理领域具有重要意义。In order to improve the quality of babies'health care,the function of babies'crying recognition was proposed to be added to the nursery box.The hardware core which consistsed of TI's digital signal processor(DSP)chip-TMS320DM643 and multimedia audio codec chip-TLC320AIC23B was performed in real-time to recognize babies'crying.Firstly,the crying signal was collected by pickups and would be handled with the audio decoder chip.Then,the sound signal was sent to the DSP chip and would be pretreated.The characteristic parameters-linear prediction coefficient(LPC)would be extracted by optimized autocorrelation function algorithm.At last,the babies'crying could be recognized accurately by using the dynamic time wraping(DTW)recognition algorithm.The result would be sent to the host computer through the serial port.The recognition accuracy of the babies'crying state reaches as high as 97.1%in actual test,this study is of great importance in the field of infant care.

关 键 词：婴儿护理 哭声识别 数字信号处理器 线性预测系数 动态时间规整算法 

分 类 号：R318[医药卫生—生物医学工程] TP391.4[医药卫生—基础医学]