The application discloses a voice
signal receiving and transmitting method and device in the technical field of voice
signal transmission, and the receiving method comprises the following steps: obtaining second fused semantic features; performing dimension
recovery, inverse normalization and splitting
processing on the second fused semantic features to obtain first transmission semantic features and second transmission semantic features; performing feature
processing on the first transmission semantic features to obtain the first transmission semantic features after feature
processing; performing dimension adjustment on the second transmission semantic features to obtain the second transmission semantic features after dimension alignment; decoding the first transmission semantic features after feature processing to obtain reconstructed audio signals with optimal evaluation voice quality; and classifying the second transmission semantic features after dimension alignment to obtain prediction scores. The application can improve the reliability and security of voice transmission tasks in a low
signal-to-
noise ratio condition while saving communication resources.