Sequential segmentation of nonstationary digital signals using spectral analysis (Q1061090)

This paper is devoted to the sequential detection of abrupt changes in spectral characteristics of digital signals, as this problem occurs for the segmentation of real signals such as speech, EEG, ECG, or geophysical signals. The limitations of a classical test are emphasized, and some new algorithms are presented. They are based upon the use of two autoregressive models and some distance measures between them, such as the log-likelihood ratio and Kullback's divergence between conditional probability laws. All these algorithms are compared both via a simulation study and from a theoretical point of view.