Trausti Kristjansson, Sabine Deligne, Peder Olsen
Accurate speech activity detection is a challenging problem in the car environment where high background noise and high amplitude transient sounds are common. We investigate a numberof features that are designed for capturing the harmonic struc-ture of speech. We evaluate separately three important characteristics of these features: 1) discriminative power 2) robustnessto greatly varying SNR and channel characteristics and 3) performance when used in conjunction with MFCC features. W epropose a new features, the Windowed Autocorrelation Lag Energy (WALE) which has desirable properties.