Single-channel speech enhancement based on joint constrained dictionary learning

Abstract To improve the performance of speech enhancement in a complex noise environment, joint constrained dictionary learning method for single-channel is proposed, which solves “cross projection” problem signals dictionary. In method, new optimization function not only constrains sparse representation noisy signal dictionary, and controls projection error on corresponding sub-dictionary, but also minimizes cross correlation between sub-dictionaries. addition, adjustment factors are introduced to balance weight constraint terms obtain more discriminatively. When applied enhancement, components can be projected onto clean sub-dictionary without being affected by makes quality intelligibility enhanced higher. The experimental results verify that our algorithm has better than based discriminative under white colored environments time domain waveform, spectrogram, global signal-to-noise ratio, subjective evaluation quality, logarithmic spectrum distance.