Paper—Convolutional Deep Neural Network and Full Connectivity for Speech Enhancement Convolutional Deep Neural Network and Full Connectivity for Speech Enhancement https://doi.org/10.3991/ijoe.v19i04.37577 Ban M. Alameri 1,2() , Inas Jawad Kadhim 3 , Suha Qasim Hadi 1 , Ali F. Hassoon 1 , Mustafa M. Abd 1 , Prashan Premaratne 4 1 Department of Electrical Engineering, Faculty of Engineering, Mustansiriyah University, Baghdad, Iraq 2 Department of Telecommunication Engineering, Malaga University, Malaga, Spain 3 Electrical Engineering Technical College, Middle Technical University, Baghdad, Iraq 4 School of Electrical and Computer and Telecommunications Engineering, University of Wollongong, North Wollongong, Australia Ban.alameri@uomustansiriyah.edu.iq Abstract—The speech signal that is received in real-time has background noise and reverberations, which have an impact on the quality of speech. Therefore, it is crucial to reduce or eliminate the noise and increase the intel- ligibility and quality of speech signals. In this study, a proposed method that is the most effective and challenging in a low SNR environment for three types of noise are removed, including washing machine, traffic noise, and electric fan noise, and clean speech is recovered. with three samples of noise which are mixed and added to the clean speech signal with a lower level of SNR value fixed at (−5, 0, 5) dBs, that noise source takes equal weights. The enhancement of the corrupted speech signal is done by applying a fully connected and convo- lutional neural network-based denoising algorithm and comparing their perfor- mance. The proposed network shows that a fully connected network (FCN) has less elapsed time than a convolutional network (CNN) while still achieving better performance, demonstrating its applicability for an embedded system. Also, the results obtained show that, overall, the CNN is better than the FCN regarding maximum coloration, PSNR, MES, and STOI. Keywords—speech enhancement, deep learning, fully connected network, convolutional network, signal-to-noise ratio (SNR) 1 Introduction Under diverse communication circumstances like speech, speech signals are contin- uously distorted by numerous noises. The effect of noise on the sound signal quality is an important issue for many communication companies due to the demand for the best quality in voice and video technology. The speech signal is hampered by many types of noise, including white noise, traffic noise, babble noise, additive noise, and channel noise [1]. Noise reduction or speech enhancement are common terms used to describe how to deal with background noise [2]. The two primary categories of 140 http://www.i-joe.org