水下机器人研究室知识库

The high frequency components of the audio signal are often truncated during the encoding processing by a lossy codec. To avoid the sound quality degradation, the high frequency components are reconstructed during the decoding processing. This paper presents a new bandwidth extension method for audio compression. Frequency components of 6.9 -13.8 kHz are added using side information at 2 kbps. A generative neural network in the GAN is used to estimate relationship between the MDCT spectrum in the high frequency part and the low frequency part, and it is evaluated by a discriminant network in the GAN to get a more natural result. On this basis, a codec system is built up. The MUSHRA experiments show that the proposed method is comparable with SBR in HE-AAC.