Vol.:(0123456789) Wireless Personal Communications https://doi.org/10.1007/s11277-019-06902-0 1 3 Speech Intelligibility Based Enhancement System Using Modifed Deep Neural Network and Adaptive Multi‑band Spectral Subtraction Tusar Kanti Dash 1,2  · Sandeep Singh Solanki 1 © Springer Science+Business Media, LLC, part of Springer Nature 2019 Abstract In contrast to the adverse environments, performances of existing speech enhance- ment algorithms do not always produce satisfactory results. In the case of worst signal to noise ratio, the processing is complicated and it may introduce signal distortions and deg- radation of intelligibility. To overcome the complexity of the existing speech enhancement algorithms, a hybrid concept for enhancing the speech quality and intelligibility is proposed in this research. The primary objectives of the research work is to increase the intelligibil- ity of the speech enhancement system that has been trained for a particular speech signal using modifed deep neural network (DNN) and adaptive multi-band spectral subtraction (AdMBSS). In this work, AdMBSS is used for enhancing the intelligibility of the speech signal using the additional phase information calculation, and fnally, hybrid DNN and Nelder Mead optimization is utilized to improve the signal quality. Experimental results explain that the proposed framework achieves improved performance in signal to noise ratio, perceptual evaluation of signal quality and minimum mean square error. Finally, per- formances are taken for the more noises like bus noise, train noise, babble noise, airport noise, station noise and exhibition noise. Keywords Speech enhancement · Speech intelligibility · Feature extraction · Optimization 1 Introduction Speech Enhancement is widely used in speech communication, speech recognition and hearing aids [1, 2]. Due to some noises in the environment like car noise and high levels of ambient noises, the mobile communication may be afected [1]. To achieve a maximum communication quality, reasonable signal enhancement is necessary * Tusar Kanti Dash tusarkantidash@gmail.com Sandeep Singh Solanki ssolanki@bitmesra.ac.in 1 Electronics and Communication Engineering, Birla Institute of Technology, Mesra, India 2 Electronics and Telecom Engineering, CV Raman College of Engineering, Bhubaneswar, India