【摘要】 针对公共环境中异常声音的检测与识别存在的强噪声干扰及检测效率低的问题,提出基于参数自适应匹配跟踪的声信号识别算法。基于粒子和种群的进化率改进粒子群参数的自适应设置并优化稀疏分解目标函数;基于自适应粒子群算法的连续集搜索特性建立连续超完备Gabor原子集,以提高最匹配优原子与声信号的匹配度并加速原子的匹配搜索;使用SVM分类器实现公共环境异常声信号的复合特征识别。实验结果表明,与已有算法相比,该算法的公共环境异常声信号的识别率最优,且对不同背景噪声具有较好的识别鲁棒性。更多还原

【Abstract】 Aiming at the problems of strong noise interference and low detection efficiency in the detection and recognition of unusual sound in public environment, a sound recognition algorithm based on parameter adaptive matching tracking is proposed. The adaptive setting of PSO parameters was improved based on evolution rate of particles and populations, and the objective function of sparse decomposition was optimized. Based on the continuous set search characteristics of adaptive PSO, a continuous super complete Gabor atom set was established to improve the matching degree of the best matching atom with acoustic signal and accelerated the matching search of atoms. The SVM classifier was used to realize the complex feature recognition of unusual sound signal in public environment. The experimental results show that, compared with the existing algorithms, the proposed algorithm has the best recognition rate of anomalous acoustic signals in the public environment and has good recognition robustness to different background noises.更多还原