Genetic Algorithm‐Based Optimization for Speech Processing Applications
R. Ramya, M. Preethi, R. Rajalakshmi · 2023
Optimization techniques are meant to solve optimization of smooth problems where it follows to find gradient of the functions. Gradients look into minima value unfortunately local minima can be a hindrance. Genetic algorithm (GA) follows biological evaluation that provides fittest solution to smooth problems and many times even to discontinue functions. GA integrated with neural network enhances its learning capabilities and input selection. This integration can be a fathom to a variety of speech processing applications, like automatic speech recognition (ASR), speech emotion recognition (SER), hate speech detection, and many other. GA plays a good role in selecting the fittest parameter set in voice activity detection, feature selection, phonetic decoding of ASR. In SER, GA improves the accuracy by using clustering-based fitness function for choosing the elementary population. Modified mutation and crossover in GA model is drawn on to solve classification problem in despise speech detection problem in social media.