【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。更多还原

【Abstract】 Aimed at applying the sound recognition technology on the vehicle type recognition in border surveillance system,a new feature extraction method was proposed based on the improved time encoded signal processing(TESP) algorithm.According to the characteristics of the vehicle sound signal,an extensional symbol table with 40 characters was designed,and then the one-dimensional S-matrix was constructed based on the symbol stream which was coded by the symbol table.Meanwhile,using the probability of occurrence of the two identical consecutive symbols,the two-dimensional A-matrix was constructed as well in order to obtain more accurate features of the signal.After that,support vector machine(SVM) and KNN were used as the classifier to recognize different vehicle types.The simulation results indicated that,comparing with the recognition rate based on the conventional TESP algorithm(51%),the recognition rates using the proposed S-matrix and A-matrix were up to 84% and 87%,respectively.Comparing with the feature extraction methods based on frequency domain and time-frequency domain analysis,the improved TESP algorithm needed less computational power and energy while providing high recognition rates.更多还原