Perspective Identification of Indian Musical Instrument Ghan and Sushir Vadya Satish Ramling Sankaye #1 , S.C. Mehrotra #2 , U.S. Tandon *3 # Department of Computer Science and Information Technology Dr. Babasaheb Ambedkar Marathwada University, Aurangabad, India * Department of Physics, College of Natural and Computational Sciences, Haramaya University, Dire Dawa, Ethiopia Abstract— A lot of work has been done globally in the area of Digital Signal Processing for the recognition and identification of speech and for the study of Musical Instruments. This study presents experimental results of Identification of two musical instruments viz. Ghan and Sushir Vadya using Linear Predictor Coefficient (LPC) features and Linear Discriminant Analysis (LDA) as the Classifiers. During the experiment, sound emanating from different Ghan and Sushir Vadya instruments has been recorded as solo notes and then its LPC feature study has been carried out. The process has been conceived and performed to find significant results of about 94.44% musical instrument sound identification from 370 sound exerts using LDA method. Keywords — Ghan Vadya, Sushir Vadya, Linear Predictor Coefficients (LPC), Linear Discriminant Analysis (LDA). I. INTRODUCTION Music is an emerging and closely related topic with wide applications like media annotation, singer identification, music transcription, structured audio coding, information retrieval, detection, classification, and separation of Musical sound Signal [1]. Musical application is especially important, since musical sounds are designed merely for human audition. The study of music signal is useful in teaching and evaluation of music. Earlier studies reveal similarities in the spectral and temporal properties of musical audio signal [2] and speech signal. Hence, several techniques developed to study speech signal are employed to study music signals as well. Musical instrument Identification is edged on classification of single note (Monophonic), more than one instrument notes at a time (Polyphonic), distinction of instruments in continuous recording or Classification of family/genre [3] [4] [5]. Although Music cannot be limited in the borders of region, relation, or nation, they are classified on the basis of the orientation of the musical instruments and its wide use in particular geographical region. According to the Natyashastra of Bharatha, there are four classes of Indian musical instruments: Tata or Tantu (stringed), Avanaddha (percussion or drums), Ghana (bells, cymbals and gongs), and Sushir (wind) [6]. The present paper discusses Sushir vadya and Ghana vadya in detail. Sushir Vadya is also known as Sushira or Aerophones. Sushira means 'hollow'. It is a musical instrument producing sound primarily by causing a body of air to vibrate, without use of strings or membranes, and without the vibration of instrument itself adding considerably to the sound. All wind instruments belong to this class. Our Study is limited specifically to Bansuri, Shehnai and Harmonium belonging to the family of Sushir Vadya. Ghana Vadya also known as Idiophones, are solid instruments which do not need any further tuning. Ghana Vadya creates sound primarily by way of the instrument vibrating itself, without the use of strings or membranes. Ghana Vadya are probably the oldest type of musical instruments. They are made to vibrate by being hit hence the name Ghana, either directly with a stick or hand or indirectly, by way of a scraping or shaking motion. Various Ghana Vadya instruments, namely Ghungaroo, Manjira, Triangle and Ghatam have been studied. This paper is organized as: Section 2 has brief collections and presentation of work most relevant to the present study. Section 3 is devoted to the LPC features and LDA method. Section 4 describes the data set used, the features, and the experiments performed to assess the performance of the proposed classifier. Finally, conclusions are drawn in Section 5. II. RESEARCH REVIEW There has been a lot of research work in the area of Music Instrument Recognition (MIR) using different features set and many classification techniques. Some of the prominent research works are discussed below: Antti Eronen and Anssi Klapuri [1] in their work Musical Instrument Recognition using Cepstral Coefficient and Temporal Features correctly recognized 94% instrument family and 80% individual instruments. They focussed on the autocorrelation sequence and then used LPC coefficient calculation with Levinson-Durbin algorithm for instrument identification. Kim, Youngmoo E., and Brian Whitman [7] proposed Singer Identification in Popular Music Recordings. The System proposed by them uses features drawn from voice coding based on LPC after segmentation prior to singer identification. Janet Marques and Pedro J. Moreno [8] proposed the classification of musical instruments using GMM and SVM Methods. The set of Features used by them were linear Satish Ramling Sankaye et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 6 (3) , 2015, 2880-2883 www.ijcsit.com 2880