昆虫鸣声信号特征参数提取的研究

Studying of Detecting Characteristic Parameters for Insect's Song

介绍了昆虫鸣声处理的时域、频域处理方法,以及它们各自的优缺点。提出谱估计法进行昆虫鸣声特征信号的提取,并简述了平均周期图法、Welch法和AR谱估计算法原理,分析了它们在昆虫鸣声特征提取时的优缺点。经试验证明,用Burg法实现昆虫鸣声的AR谱估计方法,谱线光滑,且分辨率高,能够最有效的提取昆虫鸣声的特征参数,为昆虫识别技术奠定了一定的基础。

Processing for insect＇s song in time field and frequency field are introduced in this paper, and advantages ＆ disadvantages of these methods are also introduced respectively. The way of spectral estimation is put forward, algorithm principles and advantages ＆ disadvantages of Average Periodogram, Welch and Autoregressive spectral estimation are analyzed respectively as well. Based on these methods, experimental study are carried out. The experimental results show that Autoregressive spectral estimation using Burg to estimate AR model＇ s parameters has more fluency spectral line and high frequency resolution, and it can detect characteristic parameters for insect＇s song efficiently.