Concerning an adaptive algorithm for detecting disharmonies in random sequences (Q1287237)

The paper considers the problem of detection of changes in a stochastic sequence by using filter banks. In this approach each filter represents one hypothesis. The hypotheses are compared, effectively, by computing probabilities and choosing the hypothesis (filter) having the highest probability. Changes are signaled when this results in a different filter than at previous times. The authors propose the use of a linear combination of the filters in the bank instead of processing them separately. The combination is obtained as a solution to a certain extremal problem. With the additional restriction on the coefficients to be non-negative (and to sum to one), they can be interpreted as probabilities and used for decision about changes in the model.