EURASIP Journal on Audio, Speech, and Music Processing

Table 1 The specifications of the Persian ESD and CREMA-D databases

From: Feature compensation based on the normalization of vocal tract length for the improvement of emotion-affected speech recognition

		Persian ESD			CREMA-D
		Male	Female	Total	Male	Female	Total
Number of utterances	Anger	31	31	62	671	600	1271
	Disgust	29	29	58	671	600	1271
	Fear	29	29	58	671	600	1271
	Happy	29	29	58	671	600	1271
	Neutral	90	90	180	575	512	1087
	Sad	28	28	56	671	600	1271
	Total	236	236	472	3930	3512	7442
Number of speakers		1	1	2	48	43	91
Number of sentences				90			12
Sampling rate				44.1 kHz			16 kHz

Back to article page