Feature normalization for speaker verification in room reverberation

Automatic Speaker Recognition System in Adverse Conditions — Implication of Noise and Reverberation on System Performance

“…The other method, as discussed in Section 4.2 was utilized to generate reverberation samples with image source computer simulation. A different room dimensions (3×4×2.5, 3 ), five different times (0.1, 0.5, 1, 1.5 and 2s) were set with each room. Furthermore, the distance between microphone and source were also variable and three distances were used as shown in Table II.…”

“…The noisy utterances were obtained by mixing noises with the clean speech with variable ratios according to (3). Fig.…”

“…Fig. 5 (3) where, S the is the speech signal, NR, the noise ratio and NS, is the noise signal. Nine speech samples are mixed with 3 types of noises to produce (297)speech over noise samples    4×4×2.5 and 5×4×2.5 m × × distributed in 11 mixing groups for each type of noise depends on mixing percentage.…”

“…To address this issue, various techniques have been proposed. Microphone arrays [1], score normalization [2], feature normalization [3] and alternative feature representations [4] have been suggested. The impacts of reverberation on neural network based speaker recognition has been studied in [5].…”

“…Several methods for Speech enhancement, such as spectral subtraction, have been investigated for noise-robust speaker recognition [7]. In the feature domain, for instance, techniques such as Cepstral Mean Subtraction (CMS) [1], Relative Spectral (RASTA) processing [2], and feature mapping [3] have been utilized to reduce additive and convolutional channel distortions. Previous efforts were typically reistracted to simplistic and specific case reports with too limited information to formulate the relations between the system performance and acoustic conditions, e.g.…”