应用机器听觉的木材自动识别方法研究  

Research on Automatic Wood Recognition Method Based on Machine Hearing and Support Vector Machines

作　　者：王杨 张乐 李凤妹 何昕 陈鹏 WANG Yang;ZHANG Le;LI Feng-mei;HE Xin;CHEN Peng(School of Computer and Information,Anhui Normal University,Wuhu 241000,China;Anhui Province Key Laboratory of Intelligent Building and Building Energy Saving,Anhui Jianzhu University,Hefei 230022,China)

机构地区：[1]安徽师范大学计算机与信息学院,安徽芜湖241000 [2]安徽建筑大学智能建筑与建筑节能安徽省重点实验室,合肥230022

出　　处：《小型微型计算机系统》2023年第10期2188-2195,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61871412)资助;安徽建筑大学智能建筑与建筑节能安徽省重点实验室开放项目(IBES2020KF05)资助;安徽省高等学校自然科学重点项目(KJ2020A1192)资助.

摘　　要：木材识别在智能家具制造、医药和建造等领域具有重要的应用价值.针对人类经验和视觉分析耗时费力、识别准确率不高等问题,提出了一种基于机器听觉和支持向量机的木材自动识别方法.首先,利用智能终端内置麦克风来收集声音样本数据;其次,对收集的数据进行预加重、分帧、加窗等预处理;然后对预处理后的声音数据进行特征提取,通过快速傅里叶变换(FFT)、对数运算等操作提取其梅尔倒谱系数(MFCC);最后,通过训练支持向量机(SVM)分类器来实现木材识别.实验结果表明,该方法能够以90%左右的准确率识别纤维板、夹板、柞木、橡胶木、亚克力板和胡桃木6种木材.与基于特征数据库检索和基于光谱分析的方法相比,该方法能有效的减少识别时间,提高了木材识别准确率.Wood recognition is an important application value in some intelligent furniture manufacturing,medicine and construction.Aiming at solving the problems of time-consuming and low recognition accuracy results through human experience and visual analysis,this paper proposed an automatic wood recognition method based on machine hearing and support vector machine.First,the method used the built-in microphone of the smart terminal to collect data;Second,the method performed preprocessing operations such as pre-emphasis,frame and windowing on the collected acoustic data;then the method performed feature extraction on the preprocessed data,extracting their Mel-frequency cepstral coefficient(MFCC)by fast fourier transform(FFT),logarithm operation and other operations;Finally,the method trained the support vector machine(SVM)classifier to identify the wood.The experimental resulted show that the method could identify 6 types of wood with an average accuracy of about 90%,their are fiberboard,splint,oak wood,rubber wood,acrylic board and walnut.Compared with the methods based on feature database retrieval system and spectral analysis,this method could effectively reduce the recognition time and improve the accuracy of wood recognition.

关 键 词：机器听觉 木材识别 支持向量机 声学特征 梅尔倒谱系数 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]