Search

Different from the acoustic microphone, the visual microphone uses high frame rate cameras to restore sound by using sound vibrations to cause subtle differences between frames of surrounding objects. 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 mapped the video frame image matrix with the function construct matrix. The algorithm generated chaotic attractors to preserve and amplify 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. [ABSTRACT FROM AUTHOR]