Improving speech quality in hearing aids using fuzzy complex and wavelet
Ali Keshavarz, Mohammad Divandari · 2022
This paper attempts to improve speech quality in hearing aids applications using machine learning and frequency domain analysis. For this purpose, we considered a new approach for feature extraction. This approach was using signal energy into wavelet transform for feature extraction by applying just effective features instead of sending all data to Fuzzy set. This technique on wavelet transforms was a suitable way to improve evaluation metrics and recognition time in our study. For the purpose of comparison of the proposed method, SVM and MLP have been utilized. In order to validate our method, five criteria have been used including accuracy, precision, recall, F-score, and MCC. The results of the simulation suggest that our method achieves 97% and 92% for accuracy and precision, respectively. When compared to other approaches, this shows 19% and 10% improvements in accuracy and precision with respect to SVM technique.