Genetic algorithms for component analysis
Adelino R. Ferreira da Silva · 2000
Adapted waveform analysis is used extensively in audio, speech, and video coding. In this context, the basic problem is to nd out the best representation for a given signal. The strategy is to decompose the signal into time-frequency atoms. The main analysistoolisthe expansion of the signal in orthonormal bases whose elements have good time-frequency localizations. We show that genetic algorithms can be exploited to search libraries of bases for atomic pattern recognition in noisy signals. The proposed algorithm generates best basis decomposition trees through the evolution of well-adapted genetic sequences. Two new types of constraint operators are introduced to guarantee that valid genetic sequences are generated. The approach behaves as an atomic analyzer for signal processing. A test environment is proposed to assess recognition of time-frequency atoms in signal spaces. 1