Enhanced Feature Extraction Approaches for Detection of Sound Events Naren Surampudi, Srirangan. M Department of Electrical and Electronics Engineering BITS – Pilani (Hyderabad Campus) Telangana, INDIA Jabez Christopher Department of Computer Science and Information Systems BITS – Pilani (Hyderabad Campus) Telangana, INDIA jabezc@hyderabad.bits-pilani.ac.in Abstract — This work focusses on using digital signal processing techniques to analyze and extract audio features and use them to predict the type of event that might have taken place in an audio signal using supervised machine learning approaches. The performance of five classification approaches using different feature subsets were analysed. Feature subsets include frequencies of the segmental features, frequencies of the supra-segmental features and combination of both. This gives an insight about the relative importance of the feature subsets and also the need for extracting new features from existing features. Features were extracted after the audio signal was filtered using a lowpass Butterworth filter with a cutoff frequency of 1500 Hz; it was inferred that the including features of the difference signals improved the performance of the learning algorithms. The work also includes tuning the parameters of the classification approaches to improve the performance. The observations and inferences of the experimental results can potentially be used for designing robust surveillance systems for rare event detection. Keywords— Audio Signal Processing, Rare event detection, Segmental features, Supra-segmental features, Machine learning. I. INTRODUCTION An audio signal is a form of sound energy, generally represented using a level of electrical voltage for analog forms, and binary numbers for digital forms. Digital audio systems represent audio signals in a variety of digital formats. Audio signal processing is a subfield of signal processing that is concerned with the electronic manipulation of audio signals. This involves processing signals using digital circuits such as digital processors, microprocessors and general-purpose computers. Processing approaches and use cases include acoustic detection, music information retrieval, storage, speech processing, data compression, localization, transmission, noise cancellation, acoustic fingerprinting, and sound recognition. It also includes enhancements such as equalization, filtering, level compression, echo and reverb. Audio signal processing and audio classificationiisia ipart iof ithe ilarger iproblem iof iaudio idata ihandling iwith iimportant iapplications iin idigital ilibraries, iprofessional imedia iproduction, ieducation, ientertainment iand isurveillance. The processes involved in an audio analysis system is presented in figure 1. It involves digital signal processing and machine learning. iAcoustic iscene iclassification and rare ievent idetection, ican ibe ithought iof ias ia iproduct iof ithese ifactors, iwith ione ivery icommon iuse iproposed ibyimanyipeople ibeingiusingirare ievent idetectionito idetect iany ipotential idisturbances iin ithe ineighborhood iand iaccordingly ialert ithe iusers ias iwell ias iauthorities ito ipotentially ibring idown icrime irate. iThis iis ijust ione iof ithe imany iapplications iproposed iby ia inumber iof ipeople [1-3]. iAssigning iclass ilabels ito isounds ior iaudio isegments iis iaided iby ian iunderstanding iof ieither isourceior isignal icharacteristics. Rare event detection has been one of the most researched fields in recent times, with it finding various applications in industries that deal with digital signals. One example would be the application of rare event detection to detect the occurrence of a gunshot, which enables an application to inform the user or the concerned authorities about some potential disturbance. Fig. 1. Audio Analysis System Many of the techniques and methods that have been, and are being used for rare event detection; they derive their roots from various digital signal processing techniques and machine learning algorithms. The basic digital signal processing techniques are still being extensively researched with the advent of various applications involving speech processing as well as image processing. Combining these techniques and concepts with machine learning and tuning them specifically for rare event detection opens up new avenues of research. Moreover, with the recent advent of deep learning, many researchers have begun employing more complex neural network architectures and learning paradigms for rare event detection. These methods also involves feature engineering, specifically extracting audio features, which catalyses the consequent phases of audio processing, pattern recognition and prediction. This article presents a few analysis on how feature extraction approaches can be enhanced for improving recognition and prediction of rare events. 223 978-1-7281-4392-7/19/$31.00 c 2019 IEEE