The structure of signals may be modeled by their covariance functions. Since the knowledge of a covariance function is identical to the knwledge of a spectral distibution function, we may use a spectral distribution function for analizing changes of the structure of speech signals. We can then use the analysis for discriminating their characters. Here, we use a Kalmogorov-Smirnov nonparametric hypothesis test in frequency domain. The investigation use voice and unvoice speech signals, keyboard signals dan autoregressive hypothetical models. The results show that the Kolmogorov-Simrnov method developed is able to discriminate characters in human speech signals with less than 5% significant levels.
SG Ferryanto and D. Gunawan
SG Ferryanto and D. Gunawan
