Root homomorphic deconvolution schemes for speech processing in car noise environments

The authors propose the use of a Root homomorphic approach instead of the classical Log in order to improve cepstral representation in noise. Recognition tests demonstrate the robustness of the Root-based analysis for noisy speech processing. It is shown that Root cepstral coefficients permit a heuristic extension of noise subtraction techniques into the cepstral domain. Finally, the authors extend their work to LP (linear prediction)-based cepstral analysis and show that the Root approach gives equivalent results when comparing direct (LFCC) and LP-based (LPCC) analysis schemes. A unified model is proposed, and it is concluded that both classical LPCC and LFCC analyses are suboptimal solutions of the authors´ more general scheme.<>