This paper describes a novel GMM-based bandwidth extension (BWE) method based on a sub-band basis spectrum model (SBM), in which each dimensional component represents a specific acoustic space in the frequency domain. The proposed method can achieve the BWE from a speech data with an arbitrary frequency bandwidth while the conventional methods perform the conversion from a fixed narrowband data. In the proposed method, we train a GMM with SBM parameters extracted from wideband spectra in advance. An input signal with a limited frequency band is converted into a wideband signal by estimating high-band SBM components from low-band SBM components of the input signal based on the GMM. The results of some objective and subjective evaluations show that the proposed method extends bandwidth of speech data robustly.
Bibliographic reference. Ohtani, Yamato / Tamura, Masatsune / Morita, Masahiro / Akamine, Masami (2014): "GMM-based bandwidth extension using sub-band basis spectrum model", In INTERSPEECH-2014, 2489-2493.