EURASIP Journal on Advances in Signal Processing

Table 8 Comparisons with the state of the art in terms of SIA

From: Evaluation of a speaker identification system with and without fusion using three databases in the presence of noise and handset effects

Authors	Database	System approach	Cond.1	Cond.2	Cond.3	Cond.4	Cond.5
Proposed work by		Fusion based
Al-Kaltakchi et al.	120 speakers	GMM-UBM		SNR 30 dB	SNR 30 dB	SNR 30 dB	SNR 30 dB
	Microphone channel
Proposed work by		Fusion based
Al-Kaltakchi et al.	120 speakers	GMM-UBM		SNR 30 dB	SNR 30 dB	SNR 30 dB	SNR 30 dB
Proposed work by		Fusion based
Al-Kaltakchi et al.	120 speakers	GMM-UBM		SNR 30 dB	SNR 30 dB	SNR 30 dB	SNR 30 dB
	Microphone channel
Kumar et al.	TIMIT	GMM	93.88%
[5] [2012]	120 speakers
Togneri and Pullella	TIMIT	GMM-UBM	94.5%	74.2%
[6] [2011]	64 speakers			at SNR 30 dB
Ming et al.	TIMIT	GMM	96.51%		92.86%
[51] [2007]	630 speakers	Mix 128			at 20 dB
					without handset

The colored data reflected three different databases and the highest SIA for each database: red for TIMIT, blue for SITW and Violet for NIST 2008 database. The colored italic entries represent the highest SIA

Back to article page