基于改进TESP算法的边防车辆类型声音识别

发布时间：2018-07-02 12:57

本文选题：车辆类型识别 + 实时编码信号处理(TESP)　；参考：《四川大学学报(工程科学版)》2014年S2期

【摘要】：将声音识别技术应用于边防车辆类型的自动识别,提出了一种改进的基于实时编码信号处理(TESP)算法的特征提取方法。具体方法是:根据车辆声音信号的特点提出一个包含40字符的扩展符号表,根据符号表编码后产生的符号流构造1维S矩阵,同时,以符号流中相同的连续2个符号出现的概率为参数构造2维A矩阵,得到更加精确的特征向量。然后,设计了支持向量机(SVM)分类器和K-邻近分类器(KNN)对不同的车辆类型进行分类识别。仿真实验结果表明:与基于传统TESP算法的识别率(51%)相比,基于40字符符号表的1维S矩阵和2维A矩阵特征提取方法的平均识别率分别达到84%和87%。与传统的基于频域和时频域提取特征方法相比,该算法需要较少的运算能量和内存资源,识别速率快,识别准确率高。
[Abstract]:An improved feature extraction method based on real time coded signal processing (TESP) algorithm is proposed by applying sound recognition technology to automatic recognition of border vehicle types. The specific methods are as follows: according to the characteristics of vehicle sound signals, an extended symbol table containing 40 characters is proposed, and a 1-dimensional S matrix is constructed according to the symbol stream encoded from the symbol table, and at the same time, The 2-dimensional A matrix is constructed by taking the probability of the same continuous two symbols in the symbol flow as a parameter to obtain a more accurate eigenvector. Then, support vector machine (SVM) classifier and KNN (KNN) are designed to identify different vehicle types. Simulation results show that the average recognition rates of 1-D S matrix and 2-D A matrix based on 40 character symbol table are 84% and 87%, respectively, compared with the traditional TESP algorithm (51%). Compared with the traditional feature extraction methods based on frequency domain and time and frequency domain, this algorithm needs less computing energy and memory resources, and the recognition rate is fast and the recognition accuracy is high.
【作者单位】：四川大学电子信息学院;四川大学计算机学院;
【基金】：国家自然科学基金资助项目(61272448;61302028) 四川省科技支撑计划资助项目(2012RZ0005) 四川大学青年教师科研启动项目(2012SCU11036)
【分类号】：TN912.34

【相似文献】