Abstract: A new method was proposed based on a feature fusion matrix for objective evaluation of speech timbre thickness in this paper. The feature matrix was consisted of two parts, the speech feature parameter of MFCC and LPC, and the non-speech feature of resonance factor, body mass index, and vital capacity weight index. These features were fused in different combinations to establish a feature matrix. Finally, a softmax classifier was employed to classify the timbre thickness for speech into six levels. Experimental results show that the proposed method can achieve high performance for objective assessment.