基于坐标—灰度交织映射的视频麦克风声音复原系统

区别于声学麦克风，视频麦克采用高帧率摄像机，利用声音震动引起周围物体的帧间图像细微差异，对声音进行复原。模拟人脑v2皮层位置亮度交织神经网络，提出一种基于二维对数的坐标—灰度交织映射算法。算法将视频帧图像矩阵与函数构造矩阵交织映射，产生混沌吸引子保留并放大帧间图像微差异，并构造基于该算法的视频麦克风声音复原系统。仿真实验表明，该视频麦克风系统可以从薯片袋子的高帧率视频中复原声音信号，相比于传统算法降低了40%的计算量，同时保持语音复原质量。

Visual microphone sound restoration system based on coordinate-gray interleaving mapping

Different from the acoustic microphone, the visual microphone uses high frame rate cameras to capture sound. It uses sound vibrations to cause subtle differences between frames of surrounding objects to restore sound. By simulating the positional brightness interleaving neural network of v2, this paper proposed a two-dimensional logarithm-based coordinate-gray interleaving mapping algorithm. The algorithm interleaved and maped the video frame image matrix with the function construct matrix. The algorithm generated chaotic attractors to preserve and amplified the micro-differences between frames, and constructed the visual microphone sound restoration system based on the algorithm. Simulation experiments show that the visual microphone system can recover the sound signal from the high frame rate video of the bag of chips. Compared with the traditional algorithm, this algorithm reduces the calculation amount by 40%, while maintaining the quality of speech restoration.