A robust technique for noise speech recognition is described. Based on suitably combining cepstral and group delay spectrum coefficients, a robust feature for noise-resistance of speech is proposed. In experiments, a remarkable recognition rate was obtain
