International Journal of Electrical and Electronics Research ISSN 2348-6988 (online) Vol. 8, Issue 1, pp: (9-15), Month: January - March 2020, Available at: www.researchpublish.com
Non-stationary additive noise signal filtration process in NMF based approach for the singlechannel speech enhancement 1
Ravi Shankar Prasad, 2Pradeep singh yadav
M. Tech Scholar, Assistant professor (Electronic and Telecommunication) SSITM Ravishankar.parased@gmail.com, pradeepyadav.py3@gmail.com
Abstract: This paper investigates a non-negative matrix factorization (NMF)-based approach to the semisupervised single-channel speech enhancement problem where only non-stationary additive noise signals are given. The NMF spectral basis matrices for both speech and noise are obtained in a manner of supervised learning, and thus the performance of their associated NMF speech enhancement degrades as the speaker and/or noise characteristics are not matched for the learning and evaluation environment. The experimental evaluation was made on TIMIT corpus mixed with various types of noise. It has been shown that the proposed method outperforms some of the state-of-the-art noise suppression techniques in terms of signal-to-noise ratio. Keywords: non-negative matrix factorization (NMF), noise suppression techniques, signal-to-noise ratio.
1. INTRODUCTION Signal means information and processing means operation. It means how information in the form of signal is operated or modified to get desired signal and how system process these signal[1]. Signal processing is very wide field. We are all immersed in a sea of signals. All of us from the smallest living unit, a cell, to the most complex living organism (humans) are all time receiving signals and are processing them. Survival of any living organism depends upon processing the signals appropriately. What is signal? To define this precisely is a difficult task. Anything which carries information is a signal. In this course we will learn some of the mathematical representations of the signals, which has been found very useful in making information processing systems[2]. Examples of signals are human voice, chirping of birds, smoke signals, gestures (sign language), and fragrances of the flowers. Many of our body functions are regulated by chemical signals, blind people use sense of touch. Bees communicate by their dancing pattern. Speech enhancement in presence of background noise is an important problem that exists for a long time and still is widely studied nowadays. The efficient single-channel noise suppression (or noise reduction) techniques are essential for increasing quality and intelligibility of speech, as well as improving noise robustness for automatic speech recognition (ASR) systems[3]. Generally speaking, the aforementioned techniques can be a form of machine learning, which is usually divided into two categories: supervised learning and unsupervised learning[4]. In supervised learning the training data are well classified and labeled while in unsupervised learning they are not. By processing we mean operating in some fashion on a signal to extract some useful information. For example when we hear same thing we use our ears and auditory path ways in the brain to extract the information[3]. The signal is processed by a system. In the example mentioned above the system is biological in nature. We can use an electronic system to try to mimic this behavior. The signal processor may be an electronic system, a mechanical system or even it might be a computer program. The word digital in digital signal processing means that the processing is done either by a digital hardware or by a digital computer.
Page | 9 Research Publish Journals