Research on ultrasonic communication echo elimination based on fully convolutional time-domain audio separation network

Ultrasonic communication has the characteristics of high precision positioning and anti-electromagnetic interference, but echo interference will affect its quality. In order to solve the problem that echo interference affects communication quality, an echo cancellation method of ultrasonic communication based on full convolution time domain audio separation network is proposed in this paper. The audio signal is directly processed in time domain by deep learning model to realize end-to-end learning from the original signal to the separated signal. The experimental results show that the peak signal-to-noise ratio and structural similarity index are improved after image processing based on the full convolution time domain audio separation network method. When the signal-to-noise ratio is 35dB, the peak signal-to-noise ratio of the method is the largest, reaching 28.24dB, which is significantly higher than the 16.98dB of the unprocessed image. The method effectively improves the peak signal-to-noise ratio of the image after processing, eliminates the echo of ultrasonic communication, and improves the accuracy and reliability of ultrasonic communication. The proposed method has remarkable effect in improving the image quality of ultrasonic communication, and provides an effective technical means for interference elimination in ultrasonic communication field.