Pop-Click Noise Detection Using Inter-Frame Correlation for Improved Portable Auditory Sensing Dong Yun Lee, Kwang Myung Jeon, and Hong Kook Kim School of Information and Communications Gwangju Institute of Science and Technology (GIST) {ldy, kmjeon, hongkook}@gist.ac.kr Abstract. In this paper, a pop-click noise detection method is proposed to im- prove the quality of audio signals recorded using portable microphones. In or- der to reduce false alarm and missing detection errors, the proposed method uti- lizes the second-order difference of an uncorrelated residual signal, followed by adaptive median thresholding. It is shown from performance evaluation that the proposed method achieves higher detection accuracy, under various signal-to- noise ratio conditions, than with a conventional method using the first-order dif- ference of the residual signal. Keywords: Pop-click noise, linear prediction, difference of residual signal, adaptive median thresholding 1 Introduction When an audio signal is recorded by portable auditory sensors such as a condenser microphone or a micro electrical-mechanical system (MEMS) microphone, there are many factors that generate noise. In particular, a pop-click noise is generated by vari- ous physical phenomena in acoustics, such as touching screens, clicking buttons, and so on [1]. Interference by such noises while recording can be highly annoying to most listeners. In order to detect a pop-click noise, there have been many noise detection meth- ods proposed [1–4]. Among them, the techniques using the residual signal [3] and the first-order difference of the residual signal [4] were reported to successfully detect pop-click noises. However, the error rate increased extremely in the interval where the signal-to-noise ratio (SNR) was high [4]. Thus, in order to improve detection accuracy under high SNR conditions, the pro- posed pop-click noise detection method first utilizes the second-order difference of an uncorrelated residual signal. This gives more emphasis on a pop-click noise as it has strong energy in the high frequency band. Next, adaptive median threshold [5][6] is applied at those values to determine whether or not there is a pop-click noise. Following this introduction, Section 2 proposes a pop-click noise detection meth- od using the second-order difference of an uncorrelated residual signal. Next, Section 3 evaluates the performance of the proposed method and compares it with that of a Advanced Science and Technology Letters Vol.30 (ICCA 2013), pp.164-168 http://dx.doi.org/10.14257/astl.2013.30.35 ISSN: 2287-1233 ASTL Copyright © 2013 SERSC