Articulatory-to-Acoustic Conversion of Mandarin Emotional Speech Based on PSO-LSSVM

Table 4

Recognition results on bimodal emotional speech database based on different features (%).

Feature type	Emotion type				Average rate of recognition
	Anger	Happiness	Neutral	Sadness
Original articulatory features	62.72	69.23	38.1	50	55.01
Converted F2	48.67	40.51	49.87	35.1	43.54
Converted MFCC (12-D)	53.07	41.2	50.25	38.23	45.69
Fused conversion features	58.43	47.22	54.11	38.1	49.47
True F2	54.21	53.91	56.54	52.11	54.19
True MFCC (12-D)	53.03	52.32	56.22	48.23	52.45
Fused true feature	55.46	52.98	59.54	56.11	56.02