Word Spotting in Bitmapped Fax Documents
Information Retrieval
Approximate Time-Variable Coherence Analysis of Multichannel Signals
Multidimensional Systems and Signal Processing
Robust speaker verification with optimal pitch bases expansions
WISICT '04 Proceedings of the winter international synposium on Information and communication technologies
| Hi-index | 0.00 |
Time-frequency(t-f) analysis has clearly reached a certain maturity. One cannow often provide striking visual representations of the jointtime-frequency energy representation of signals. However, ithas been difficult to take advantage of this rich source of informationconcerning the signal, especially for multidimensional signals.Properly constructed time-frequency distributions enjoy manydesirable properties. Attempts to incorporate t-f analysis resultsinto pattern recognition schemes have not been notably successfulto date. Aided by Cohen‘s scale transform one may construct representationsfrom the t-f results which are highly useful in pattern classification.Such methods can produce two dimensional representations whichare invariant to time-shift, frequency-shift and scale changes.In addition, two dimensional objects such as images can be representedin a like manner in a four dimensional form. Even so, remainingextraneous variations often defeat the pattern classificationapproach. This paper presents a method based on noise subspaceconcepts. The noise subspace enhancement allows one to separatethe desired invariant forms from extraneous variations, yieldingmuch improved classification results. Examples from sound classificationare discussed.