Development of the Arabic Voice Pathology Database and Its Evaluation by Using Speech Features and Machine Learning Algorithms

Table 7

Overall best accuracies (%) for sustained vowels and running speech by using the AVPD.

Features	Experiments	SVM		GMM		VQ		HMM
		/AH/	Al-Fateha	/AH/	Al-Fateha	/AH/	Al-Fateha	/AH/	Al-Fateha
MFCC	Detection		77.4	74.4	77.1	70.3	71.1	71.6
	Classification		89.2	88.9	89.5	75.3	81.6	88.7
LPCC	Detection	60.1	76.5	54.5		70.3	75.9		71.5
	Classification	67.6	84.7	75.4			77.9	59.0
RASTA-PLP	Detection		76.7	72.8	74.5	67.1	75.0	66.3
	Classification		90.2	91.3	91.2	88.9	90.3	88.7
LPC	Detection	62.3	71.6	53.7		70.7	71.5		62.3
	Classification	66.3		74.6	79.7	78.6	75.3		75.9
PLP	Detection		79.1	73.2	78.5	72.0	78.1	73.6
	Classification		90.1	88.9		79.4	77.2	88.7	85.8
MDVP	Detection		—	69.8	—	64.8	—	—	—
	Classification		—	—	—	—	—	—	—

best detection rate for sustained vowels. The best detection rate for running speech. The best classification rate for sustained vowels. The best classification rate for running speech.