摘要
本文提出一种声源触发端点识别检测算法,以解决在智能交通系统中复杂噪声环境下不能有效提取鸣笛信号的问题。该算法利用倒谱分析的思想,分析并提取接收信号短时频域特征,准确判断声源信号的起止端点。该方法不依赖于统计鸣笛信号与背景噪声的先验知识,也不需要多次测试设定经验阈值。实际数据验证表明该方法能准确识别出一段信号中声源触发起始帧。
This paper presents an algorithm of detecting the initial endpoint of audio source,solving the problem that the whistle signal cannot be seperated effectively in the complicated noise circumstances.In the algorithm,we analyse and ectract the short-time frequency domain properties of the accepted signal based on the idea of Cepstrum analysis,then find the starting terminal of the triggered signal accurately.The algorithm relies on neither the prior knowledge of whistle signal and background noise,nor the threshold set from the experienced tests.Moreover,simulation results utilizing actual data indicate that the algorithm can identify the triggered frame of a section mixed signals precisely.
作者
祁兵兵
刘金磊
张焕胜
侍艳华
田晟昊
窦硕鹏
姜孝超
QI Bingbing;LIU Jinlei;ZHANG Huansheng;SHI Yanhua;TIAN Shenghao;DOU Shuopeng;JIANG Xiaochao(Beijing China Electronics Intelligent Acoustics Technology Co.,Ltd.,Beijing 100015,China)
出处
《电声技术》
2018年第10期8-13,68,共7页
Audio Engineering
关键词
端点检测
鸣笛信号
倒谱分析
endpoint detect
whistle signal
Cepstrum analysis
