Wavelet analysis is a powerful tool for analyzing speech and acoustic signals, particularly useful when speech intelligibility is compromised by factors like face masks. The COVID-19 pandemic highlighted the challenges masks pose to clear communication. Wavelet Transform (WT) offers a solution by integrating time and frequency domain data to improve speech recognition.
The selection of an appropriate "mother wavelet" is critical for effective WT, as different wavelets yield varying results. Research leverages the COPRAS (COmplex PRoportional ASsessment) technique to determine the optimal mother wavelet function for speech signals when face masks or shields are used.
Metrics like Maximum Cross-Correlation Coefficient (MCC) and Maximum Energy to Shannon Ratio (MEER) are employed to rank the mother wavelet functions. This method establishes a clear protocol for selecting the most suitable mother wavelets for speech signals in diverse, real-world conditions where masks may be present. Wavelet transforms improve pattern recognition systems by extracting features that are invariant to certain transformations and can improve the performance of classifiers in noisy environments.