一种强噪声干扰下的炮控系统声音识别算

A sound recognition algorithm for gun control system under the interference of strong noise

摘要:

声音作为一种重要的信息媒介，能够为维修人员提供大量的装备信息。但实际维修环境受到车辆启动噪声的干扰，难以准确直观的对声音进行判断。为实现对炮控系统各主要声音部组件启动过程的识别，提出了一种基于改进谱减法降噪和多类型识别策略的声音识别算法。通过对炮控系统各部组件与发动机声音信号的分析，利用改进谱减法对声音样本进行了降噪处理，并通过实验优化了谱减参数，进一步提升了降噪性能，解决了强噪声干扰的问题。利用滑窗校正和短时能量同步检测的方法制定了具体的识别策略，解决了实际应用中识别结果不稳定以及多类型过程识别的问题。通过实验验证，该声音识别算法对炮控系统各部件启动状态识别准确率达92.4%，具有较好的识别性能。

Abstract:

However, the actual maintenance environment is disturbed by vehicle starting noise, so it is difficult to accurately and intuitively judge the sound. In order to recognize the start-up process of the main voice components of gun control system, a voice recognition algorithm based on improved spectral subtraction noise reduction and multi-type recognition strategy was proposed. Through the analysis of the sound signals of gun control system components and engine, the improved spectral subtraction method is used to reduce the noise of sound samples, and the spectral subtraction parameters are optimized through experiments, which further improves the performance of noise reduction and solves the problem of strong noise interference. A specific recognition strategy is developed by using sliding window correction and short-time energy synchronization detection, which solves the problems of unstable recognition results and multi-type process identification in practical application. Experiments show that the recognition accuracy of this algorithm is 92.4% for the start-up status of each component of gun control system, and it has good recognition performance.