Subband DCT and EMD Based Hybrid Soft Thresholding for Speech Enhancement

Table 2

Comparison of overall SNR, average segmental SNR (AvgSegSNR), and PESQ improvements of different denoising methods for pink and HF channel noise.

Input SNR (dB)			Output SNR (dB)
	0	5	10	15	25	30
PINK
WP [3]	2.57	7.19	11.66	15.81	22.69	25.20
DCT [11]	2.12	6.78	11.35	15.81	24.58	28.98
S. DCT [4]	1.41	5.98	10.73	15.51	25.24	30.13
𝑈EMD	4.51	8.27	12.41	16.81	26.01	30.44
HF
WP [3]	1.96	6.72	11.63	16.45	24.24	26.47
DCT [11]	3.59	7.84	11.88	15.94	24.11	28.21
S. DCT [4]	0.94	5.38	10.08	14.92	24.70	29.61
𝑈EMD	4.92	8.95	12.96	17.14	26.21	30.84
In. AvgSegSNR (dB)			Output AvgSegSNR (dB)
	−4.047	−1.124	2.256	5.959	14.059	18.188
PINK
WP [3]	−2.983	0.017	3.196	6.373	12.354	14.904
DCT [11]	−3.149	−0.162	3.057	6.435	13.695	17.526
S. DCT [4]	−3.598	−0.649	2.704	6.328	14.292	18.341
𝑈EMD	−1.594	0.927	3.538	7.074	15.088	18.834
In. AvgSegSNR (dB)			Output AvgSegSNR (dB)
	−4.162	−1.287	2.079	5.781	13.906	18.049
HF
WP [3]	−3.574	−0.476	3.006	6.685	13.441	16.017
DCT [11]	−2.683	0.218	3.219	6.411	13.319	17.007
S. DCT [4]	−4.171	−1.349	1.948	5.599	13.603	17.725
𝑈EMD	−1.234	1.526	4.416	7.671	15.342	19.239
Input SNR (dB)			PESQ
	0	5	10	15	25	30
PINK
Input	1.33	1.68	2.06	2.43	3.22	3.61
WP [3]	1.64	2.04	2.38	2.66	3.15	3.32
DCT [11]	1.91	2.27	2.59	2.93	3.51	3.77
S. DCT [4]	1.85	2.17	2.51	2.84	3.50	3.79
𝑈EMD	1.93	2.29	2.62	2.95	3.55	3.83
HF
Input	1.58	1.84	2.14	2.44	3.15	3.49
WP [3]	1.67	1.87	2.12	2.45	3.15	3.47
DCT [11]	1.60	1.83	2.13	2.46	3.11	3.37
S. DCT [4]	1.49	1.62	1.84	2.14	2.94	3.32
𝑈EMD	1.61	1.96	2.32	2.66	3.34	3.65