In this paper, a blind bandwidth extension method of audio signals is proposed in which the fine structure of high-frequency information is recovered based on Volterra series. Combining with Gaussian mixture model and codebook mapping to adjust the spectrum envelope and energy gain of the extended high-frequency components separately, the bandwidth of audio signals is extended to super-wideband from wideband. Furthermore, the proposed method is applied into a real audio codec. The performance of the proposed method is evaluated through objective and subjective tests on the audio signals selected from MPEG items, and it is found that the proposed method outperforms the chaotic prediction method and nearest-neighbor matching method. When the proposed algorithm is applied into ITU-T G.722.1 wideband audio codec, the performance is comparable with that of G.722.1C super-wideband audio codec at 24 kbps.
展开▼
Speech and Audio Signal Processing Laboratory, School of Electronic Information and Control Engineering, Beijing University of Technology, 100124, China;