基于深度学习算法的激光散斑音频信号提取与增强  

Extraction and Enhancement of Laser Scattering audio Signal Based on Deep Learning Algorithm

作　　者：沃英达 WO Yingda(Inner Mongolia Autonomous Region Radio and Television Transmission and Launch Center Hailaer 548,Hulunbuir 021000,China)

机构地区：[1]内蒙古自治区广播电视传输发射中心海拉尔548台,内蒙古呼伦贝尔021000

出　　处：《电声技术》2025年第2期52-54,77,共4页Audio Engineering

摘　　要：利用深度学习算法学习激光散斑特征与音频信号参数之间的对应关系,进一步提高音频信号提取和增强的准确性。实验结果表明,在激光散斑音频信号提取与增强过程中,蚁群算法的收敛速度比深度学习算法慢,且深度学习算法最终测试损耗最低;均方根误差与信噪比变化趋势相反,深度学习算法的均方根误差仅为0.16,远低于其他算法。因此,基于深度学习算法可以有效提高激光散斑音频信号的提取与增强能力。The deep learning algorithm is used to learn the correspondence between laser scattering features and audio signal parameters to further improve the accuracy of audio signal extraction and enhancement.The experimental results show that the convergence rate of ant colony algorithm is slower than that of deep learning algorithm in the process of laser scattering audio signal extraction and enhancement,and the final test loss of deep learning algorithm is lowest.The root mean square error is contrary to the change trend of signal-to-noise ratio,and the root mean square error of deep learning algorithm is only 0.16,much lower than other algorithms.Therefore,the algorithm based on deep learning can effectively improve the extraction and enhancement ability of laser scattering audio signal.

关 键 词：深度学习算法 激光散斑 信号提取 

分 类 号：TN912.3[电子电信—通信与信息系统] TN249[电子电信—信息与通信工程] TP18[自动化与计算机技术—控制理论与控制工程]