Fig. 3From: Feature compensation based on independent noise estimation for robust speech recognitionWord error rates of the proposed algorithm with different numbers of Gaussian componentsBack to article page