Intra-frame cepstral sub-band weighting and histogram equalization for noise-robust speech recognition

Table 13 Recognition accuracy results (%) of WS-HEQ for different SNR condition at the pink noise environment

Method		Clean	20 dB	15 dB	10 dB	5 dB	Average
MFCC		76.38	59.44	44.24	22.34	5.85	32.97
CHN		76.52	61.40	52.71	38.06	24.30	44.12
${WS-HEQ}_{I}^{(1)}$	α=1.0	77.15	62.83	53.96	39.62	25.09	45.38
	α=0.6	76.94	63.48	54.99	40.76	27.10	46.58
${WS-HEQ}_{I}^{(2)}$	α=1.0	77.50	63.62	54.45	39.79	25.91	45.94
	α=0.6	76.07	63.11	55.22	40.35	26.94	46.41
${WS-HEQ}_{I}^{(3)}$	α=1.0	77.08	61.38	51.89	38.06	24.44	43.94
	α=0.6	76.84	62.55	54.34	40.60	26.61	46.03
${WS-HEQ}_{I}^{(4)}$	α=1.0	76.91	62.38	53.15	38.43	23.81	44.44
	α=0.6	76.59	63.15	54.50	40.14	26.14	45.98
${WS-HEQ}_{II}^{(1)}$	α=1.0	77.96	63.04	54.15	39.83	25.63	45.66
	α=0.6	76.54	63.62	55.27	41.51	26.84	46.81
${WS-HEQ}_{II}^{(2)}$	α=1.0	77.31	63.27	54.24	39.65	25.75	45.73
	α=0.6	76.33	62.62	55.29	40.81	26.82	46.39
${WS-HEQ}_{II}^{(3)}$	α=1.0	77.92	62.50	53.19	39.86	24.98	45.13
	α=0.6	77.01	63.48	54.90	41.02	27.08	46.62
${WS-HEQ}_{II}^{(4)}$	α=1.0	77.10	62.55	52.80	38.18	24.91	44.61
	α=0.6	77.24	63.13	54.52	40.88	26.91	46.36

These recognition accuracy results (%) of the MFCC baseline, CHN, and eight forms of WS-HEQ are for different SNR conditions at the pink noise environment as to the subset of the TCC database.