Enhancing the magnitude spectrum of speech features for robust speech recognition

EURASIP Journal on Advances in Signal Processing

Table 1 The averaged recognition accuracy rates (%) of the development set (the multi-condition training data in the AURORA-2 database) achieved by MSE with different filter coefficients λ in Equation ( 12) and exponents α in Equation (18)

The exponentα in Equation (18)	The filter coefficientλ in Equation (12)
	0.1	0.3	0.5	0.7	0.9
0.1	88.22	88.44	88.73	88.96	88.90
0.3	89.21	89.49	89.71	90.03	89.54
0.5	89.53	89.66	89.86	90.38	89.78
0.7	89.98	89.94	90.21	90.28	89.54
0.9	89.93	89.80	90.06	90.27	89.31