Abstract: Aiming at the problem that unlawful cutting of forest becomes more and more rampant at present, and there is no effective solution for real-time detection of such behavior, an approach based on audio recognition is proposed about unlawful cutting of forest. By the analysis of spectral characteristics of the sound signal, and calculation of similarity value and signal-to-noise ratio in the proposed approach, it can detect whether there is a chainsaw logging behaviour. Experimental results show that the proposed method can eliminate interference sound effectively and identify the chainsaw logging sound accurately in real time.

Key words: audio recognition, characteristics analysis, similarity value, signal-to-noise ratio

摘要： 针对目前森林盗伐猖獗，且尚无实时检测盗伐行为方法问题，提出了一种基于声音识别的森林盗伐检测方法。通过对声音信号的频谱特征分析、相似度值及信噪比计算，检测是否存在链锯伐木行为。实验结果表明，提出的方法能够有效地排除干扰声音，准确、实时地识别链锯伐木声音。

关键词: 声音识别, 特征分析, 相似度值, 信噪比