摘要
将统计数据驱动方法应用于冲击声的固有结构建模,研究了单个冲击信号及冲击声组的时域和频域特征提取方法,并将所提特征用于圆柱壳冲击声和公路空洞检测冲击声分类,获得了良好效果,证实了该方法用于冲击声目标分类的有效程度。
The statistical data-driven method is used to model the intrinsic structures of impact sounds. The method forextraction of essential features in time and frequency domains of single impact sound and ensembles of similar impactsounds are studied based on the temporal and spectral base functions. Using this method, the temporal and spectral featuresof sounds from an impacted cylinder hull and road cavity examination are extracted and then classified by an RBF-basedclassifier. It is shown that such a feature extraction method can achieve good classification performances for both types ofsounds.
出处
《噪声与振动控制》
CSCD
2014年第1期150-154,172,共6页
Noise and Vibration Control
基金
国家自然科学基金资助课题(基金号:11074202)
关键词
振动与波
数据驱动
冲击声
固有结构建模
特征提取
目标分类
vibration and wave
statistical data-driven method
impact sounds
intrinsic structure modeling
featureextraction
target classification
