计算机应用与软件Issue(1)：154-157,161,5.DOI:10.3969/j.issn.1000-386x.2014.01.040
结合时-频纹理特征的随机森林分类器应用于鸟声识别
APPLYING RANDOM FOREST CLASSIFIER COMBINED WITH TIME-FREQUENCY TEXTURE FEATURES TO BIRD SOUNDS RECOGNITION
摘要
Abstract
An anti-noise bird sounds recognition system based on time-frequency texture features and random forest classifier is studied in this paper.As various unpredictable noises existing in the practical environment can make the performance of the recognition system degrade seriously,an audio frequency enhancement algorithm based on noise-estimation is firstly applied for front-end processing of the noisy bird sounds.Then the enhanced signal power spectrum is outputted in the form of time-frequency graph,and the texture features extraction is con-ducted using texture analysis method based on gray level co-occurrence matrix (GLCM)according to the texture information contained in time-frequency graph.Finally,an ensemble classifier based on decision tree-random forest is used for the classification and recognition of bird sounds.Experimental results show that the proposed method can recognise the bird sounds quickly and accurately,and has better anti-noise property as well.关键词
生态声音/音频增强/时频纹理/灰度共生矩阵/随机森林分类器Key words
Ecological acoustics/Audio frequency enhancement/Time-frequency texture/Gray level co-occurrence matrix/Random forest classifier分类
信息技术与安全科学引用本文复制引用
陈莎莎,李应..结合时-频纹理特征的随机森林分类器应用于鸟声识别[J].计算机应用与软件,2014,(1):154-157,161,5.基金项目
国家自然科学基金项目（61075022）。 （）
