Voice Activity Detection in Noisy Environments Based on Double-Combined Fourier Transform and Line Fitting

Table 1

Comparison of the proposed algorithm to conventional algorithm in the utterance based speech segment detection test for various noise environments.

		Proposed	Sohn et al. [10]	Ramírez et al. [14]	Górriz et al. [15]	Li et al. [17]	Fukuda et al. [18]
Environments		(%)	(%)	(%)	(%)	(%)	(%)
Noise	SNR
Babble	High	92.5	89.5	92.3	93.9	85.4	90.1
	Low	88.7	81.4	90.1	91.3	74.3	86.4
Restaurant	High	91.4	90.1	91.9	92.2	83.2	89.8
	Low	88.5	82.4	88.4	89.2	70.0	84.5
Exhibition hall	High	95.8	92.0	94.0	94.0	90.7	94.4
	Low	92.9	84.3	90.1	90.5	78.9	90.1
Car	High	95.9	91.5	94.5	94.5	87.5	94.1
	Low	92.5	86.4	90.5	91.0	77.4	88.8
Average		92.3	87.2	91.5	92.1	80.9	89.8