A Probabilistic Analysis of the Ratio Spectrum

tic signal. The discrete-time ratio spectrum of X ( k ) is Recently, Lam and Harris have developed a novel specN-I tral representation called the ratio spectrum [l]. The I K r " ( k ) 1 2 (1) Output Power k=O N-1 ratio spectrum is formed by taking the ratio of the power of a low-pass filtered signal to the power of the original unfiltered signal for all filter cutoff frequencies. FeaX(m) = Input Power IX(k)I2 tures extracted from the ratio spectrum have proven to k=O be promising for phoneme recognition 12, 31 and speech compression 14, 51. I n this work we have shown that the ratio spectrum feature sets are competitive with standard feature sets used for automatic speech recognition. To understand the underlying properties of the ratio spectrum that contribute to the generation of a useful feature set, we derived the analytic expressions that describe the behavior of the ratio spectrum for a white noise model.