The results of these analyses suggest that predicting segmentation boundaries in auditory streams by using local estimators of information content may result in calculated segmentations of temporal structure that mimic those perceived by human listeners. An algorithm estimating information content using ratios of entropies could be implemented in a computationally efficient manner since the algorithm only requires two log base 2 operations, one divide and a few lookups. The algorithm has the potential for parallel implementation in that the components for each feature size are calculated independently and then linearly combined.
The Local Information Model predicts that the probability of a temporal segmentation occurring at any point is directly related to the amount of information in the auditory stream at that point in time. This prediction is difficult to verify without a specific measurement model for estimating the local information in a stream on a moment by moment basis. Finding a reliable estimator for local information in complex stimuli remains an open problem, but a problem which if solved seems likely to hold rewards for research in auditory scene analysis.