Research Area:  Machine Learning
Speech enhancement is the process of treating noisy speech signals to improve human perception as well as improve system understanding of the signal. To keep the signal undistorted while also reducing noise is a difficult task, which results in a limited performance of speech enhancement systems. For speech signals having medium or high signal to noise ratio value, the aim is to come up with subjectively practical signals, and for signals with low SNR, the aim is to reduce noise level while still retaining the intelligibility. Many noise reduction algorithms improve overall speech quality but little progress has been made to improve speech intelligibility. In this paper the necessity of speech enhancement, its different applications, overview of classification and various methods associated with it has been presented and a substantial literature review on such speech enhancement systems with various methods and platforms is done. Deep convolution neural network-based speech enhancement system is intended by optimizing the loss functions like Extended Short-Time Objective Intelligibility and Mean Square Error. The loss function required for training it are optimized using Harris Hawk Optimization.
Keywords:  
Speech Enhancement
Deep Neural Network
Deep Learning
Machine Learning
Author(s) Name:  Ramesh Nuthakki, Payel Masanta & T. N. Yukta
Journal name:  
Conferrence name:  ICCCE
Publisher name:  Springer
DOI:  10.1007/978-981-16-7985-8_2
Volume Information:  2021 pp 7–16
Paper Link:   https://link.springer.com/chapter/10.1007/978-981-16-7985-8_2