Nonlinear Dynamic Feature Extraction Based on Phase Space Reconstruction for the Classification of Speech and Emotion

Table 4

Recognition results of the two features for different SNR (dB).

Word no.	Feature type	SNE (dB)
		15	20	25	30	Clean	Average
10	MFCC	91.90	91.90	91.90	93.81	94.76	92.85
	NLD-1 + MFCC	93.33	94.29	93.33	94.29	95.71	94.19
	NLD-2 + MFCC	93.33	94.29	93.33	94.29	95.71	94.19
	NLD + MFCC	92.85	95.24	92.38	95.24	95.71	94.28
20	MFCC	91.19	93.38	91.67	92.86	92.14	92.25
	NLD-1 + MFCC	92.38	92.81	93.57	93.10	92.86	92.94
	NLD-2 + MFCC	91.19	93.81	93.57	92.86	92.61	93.04
	NLD + MFCC	92.14	93.81	93.10	94.05	92.62	93.14
30	MFCC	87.62	89.68	91.27	92.22	89.05	90.32
	NLD-1 + MFCC	90.79	90.79	91.27	92.63	89.79	91.54
	NLD-2 + MFCC	87.62	91.11	92.06	92.22	90.79	90.76
	NLD + MFCC	88.73	91.11	92.06	92.63	91.27	91.16
40	MFCC	86.90	88.69	90.31	90.00	88.33	88.85
	NLD-1 + MFCC	87.14	90.35	91.31	90.36	88.69	89.57
	NLD-2 + MFCC	87.14	90.35	91.79	90.36	88.69	89.67
	NLD + MFCC	87.86	89.88	90.36	90.24	88.93	89.45
50	MFCC	84.29	87.71	88.95	78.76	85.82	85.11
	NLD-1 + MFCC	85.81	87.90	88.48	78.81	85.05	85.21
	NLD-2 + MFCC	85.52	87.52	89.05	78.76	85.52	85.27
	NLD + MFCC	85.52	87.52	89.05	88.10	87.71	87.58