用Bark频谱投影识别低信噪比动物声音  

Identifying low-SNR animal sounds based on Bark spectral projection

作　　者：黄鸿铿 李应[1] HUANG Hongkeng;LI Ying(College of Mathematics and Computer Science,Fuzhou University,Fuzhou 350116,China)

机构地区：[1]福州大学数学与计算机科学学院,福建福州350116

出　　处：《智能系统学报》2018年第4期610-618,共9页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(61075022);福建省自然科学基金项目(2018J01793)

摘　　要：复杂环境声影响低信噪比动物声音的自动识别。为解决这一问题,本文提出一种不同声场景下低信噪比动物声音识别的方法。该方法把声音信号进行Bark尺度的小波包分解,再使用分解系数生成重构信号的频谱,并对频谱进行投影生成Bark频谱投影特征,通过随机森林分类器实现低信噪比动物声音的识别。该文分别在流水声环境、公路环境、风声环境和嘈杂说话声环境下,以不同的信噪比,对40种动物声音进行识别实验。结果表明,结合短时谱估计法、Bark频谱投影特征和随机森林的方法对不同信噪比的各种环境声音中动物声音的平均识别率可以达到80.5%,且在–10 d B的情况下依然保持平均60%以上的识别率。In this paper, we consider the influence of complex background environments on the automatic recognition ofanimal sounds with low signal-to-noise ratios （SNRs）. We propose a method for identifying low-SNR animal sounds invarious background environments. In this method, the sound signal is decomposed by a Bark scale wavelet packet, andthe decomposition coefficient is used to generate a spectrogram of the reconstructed signal, which is projected onto aspectrogram to generate a Bark spectral projection （BSP） feature. Random forests （RF） are then used to identify animalsounds with low SNRs. We classified 40 common animal sounds with different SNRs in noise environments such asflowing water, highway, wind, and loud speech. The experimental results show that by combining the proposed meth-ods of short-time spectrum estimation, BSP, and RF in various background environments with different SNRs, the meanidentification rate for animal noises can reach 80.5%. In addition, a recognition rate above 60% can be maintained evenat –10 dB.

关 键 词：声音信号 自动识别 小波包变换 随机森林 环境声音 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]