"TiMerS: Time-Based Music Recommender System"
- 1. SVSI- 2017 organized by : CSE Department, MCKVIE sponsored by: CSI-MCKVIE Student Chapter and MCKV Institute of Engineering TiMerS:Time-Based Music Recommender System Gobinda Karmakar, Kehkasha Naushad,Sudipta Chakrabarty Dept. of MCA,Techno India, Salt Lake, EM 4/1, Sector –V, Kolkata – 700091, India
- 2. Introduction Music recommendation system enhances personalized music classifications that create a profile with the service and build up a music library based on the choice preferences using mobile cloud services. Music Recommendation System estimates the user’s music preference from a localized music collection and then generates a set of songs based on these estimates from a wider music collection. In this paper we proposed a method that generates the song library depends on the different time slots of a day.
- 4. ProposedWork Time slot based music recommender is very effective for mobile music listeners. The primary objective of this work is to find out the playing time of any song from listener’s music library and create time slot based different music libraries from generating Raga name of that particular song and then matching with the raga time playing database.
- 5. ProposedWork Contd… To find the raga name of music, it requires that the music run throughWave Surfer software which is used in this experiment to get the pitch value of that song. WaveSurfer is an open source tool for sound visualization and manipulation. WaveSurfer is free audio editing software widely used for interactive display of sound waveforms, spectral sections, spectrograms, pitch contour, and pitch transcriptions. WaveSurfer can read and write a number of pitch transcription file formats used in speech processing research. ‘.wav’ music file is appropriate to create the pitch file of the music.
- 6. ProposedWork Contd… Keeping sample encoding at line 16 and sample rate at 22050 the pitch contour of the music is generated. This will give all the pitches that are used in the song and the pitch data are saved in the .f0 format. It consists of huge number of frequencies of monotonic song and we convert this .f0 into .xls format. Then analyse the FFT (Fast Fourier Transform) of the fo file values. An FFT of a time domain signal takes the samples and gives the frequencies, amplitudes, and phases of the sine waves that make up the sound wave analyzed. In this case, the sample rate is 44.1 kHz and the FFT size is 1024, so the bin width is the Nyquist frequency (44,100/2 = 22,050) divided by the FFT size or near about 22 Hz. •
- 7. ProposedWork Contd…. Accepted only those frequency values within 50 to 500. Then the number of occurrences has been calculated of each frequency ranging from 50 to 500. The frequency that contains highest occurrences is the most important note and got the name of the most important note by matching with their fundamental frequency range. The fundamental frequency range is given in the Table 1. After calculating the most important Note has been found out the other 11 notes by using the Note Frequency ratio table and also finds out the Octave.The Note Frequency Ratio is also given inTable 1.
- 8. ProposedWork Contd…. Western Notes Indian Notes Frequency Range Note Frequency Ratio A (Dha) 215.5 – 226.5 1.000 A# (ni) 226.6 – 239.9 1.054 B (Ni) 240.0 – 252.5 1.125 C (Sa) 252.6 – 269.5 1.186 C# (re) 269.6 – 285.3 1.253 D (Re) 285.4 – 302.3 1.333 D# (ga) 302.4 – 320.3 1.415 E (Ga) 320.4 – 339.3 1.500 F (ma) 339.4 – 359.4 1.580 F# (MA) 359.5 – 380.9 1.691 G (Pa) 381.0 – 403.5 1.777 G# (dha) 403.6 – 427.5 1.893 Table 1. Frequency Range and Note Frequency Ratio of different Notes
- 9. ProposedWork Contd…. Then matching the frequency values of all the 12 notes from the Excel Sheet and if the notes are in the Excel Sheet then takes the value as one note. From each of the range of frequencies of notes, the most frequently occurring frequencies with significant amplitude among the observed frequencies are identified. In all there will be 12 such frequencies and from among these 12 the most prominent 7 frequencies are identified. We have a Raga Knowledge base that consists of 7 note structures of each Raga and we match that particular 7 note structures of our input song with the knowledge base and if found then print the name of the Raga, otherwise print Mismatch.
- 10. ProposedWork Contd…. After finding the raga name of that music, it requires to match with the raga playing time database and saving the music in preference time folders and create personalized time slot based music recommendation.
- 11. Work Flow
- 12. Result Set Analysis Test Case: SongTitle: “Aji rooth kar” Playback Singer: Lata Mangeshkar Film Name: Arzoo (1965) Music Director: Shankar Jaikishan Lyrics: Hasrat Jaipuri Cast: Rajendra Kumar, Sadhana Shivdasani, Feroz Khan, Nazima, Nasir Hussain, Film Director: Ramanand Sagar
- 13. Result Set Analysis Contd…. Firstly .wav file is used to create the pitch file of the song. Keeping sample encoding at line 16 and sample rate at 22050 the pitch contour of song is generated. All the pitches that are used in the song and the pitch data are saved in the excel sheet saved in .f0 format. Convert this .f0 file into .xls file that shown in Table 2.
- 14. Result Set Analysis Contd…. Notes Number of Occurrences Frequency Range A/Dha 1725 215.5 – 226.5 A#/Ni 4205 226.6 – 239.9 B/Ni 1770 240.0 – 252.5 C/Sa 2241 252.6 – 269.5 C#/re 0002 269.6 – 285.3 D/Re 4646 285.4 – 302.3 D/#ga 2836 302.4 – 320.3 E/Ga 1017 320.4 – 339.3 F/ma 2319 339.4 – 359.4 F#/MA 2187 359.5 – 380.9 G/Pa 4512 381.0 – 403.5 G#/dha 2993 403.6 – 427.5 Table 2. Note Occurrences of the song “Aji rooth kar”
- 15. Result Set Analysis Contd…. From Table 2 the song title “Aji rooth kar” has only 2 occurrences has been admired of C# and that is negligible and other 11 notes has very high occurrences. Therefore C# is absent in this song. The highest occurrence of frequency value 4646 and it belongs to the fundamental frequency range of “D/Re” note. Therefore D or Re is the most important note (Vadi Swar). The second highest occurrence of frequency value 4512 and it belongs to the fundamental frequency range of “G/Pa” note. Consequently G or Pa is the Second most important note (Samvadi Swar). Then we find out all the other notes using note-frequency ratio Table 3. and mapping with the values of .xls file and got the other notes as Sa/C, ma/F, Pa/G, Ni/B, ni/A#, Dha/A, Pa/G, ma/F, Ga/E, Re/D, Sa/C.
- 16. Result Set Analysis Contd…. Other Informations: Most Important Note orVadi: D/Re Second Most Important Note or Sambadi: G/Pa Ascent or Aroho: Sa/C, Re/D, ma/F, Pa/G, Ni/B Decent or Aboroho: ni/A#, Dha/A, Pa/G, ma/F, Ga/E, Re/D, Sa/C Raga Name: Desh Species or Jati: Pentatonic – Heptatonic (5 – 7)
- 17. Conclusion This work describes time slot based music recommendation system; a technique that allows listeners to personalized music depends on the appropriate raga playing time using raga time knowledgebase. This article has been implemented as a step towards developing a method to assist and evaluate the Raga recognition of Indian Music and also identify the playing time of that raga. In this contribution, there are eighty Indian song music have been tested and find out the raga name and raga playing time and create appropriate mobile music recommendation system using different time slots of a day on mobile cloud.
- 18. References Sudipta Chakrabarty, Md Ruhul Islam, Debashis De, “Modelling of Song Pattern Similarity using Coefficient of Variance” , International Journal of Computer Science and Information Sequrity (ISSN 1947-5500), pp. 388-394, 2017. Samarjit Roy, Sudipta Chakrabarty, Debashis De, “Time-Based Raga Recommendation and Information Retrieval of Musical Patterns in Indian Classical Music using Neural Network”, IAES International Journal of Artificial Intelligence (IJ-AI) (ISSN: 2252-8938), pp. 33-48, 2017. Sudipta Chakrabarty, Samarjit Roy, Debashis De, Edited by Siddhartha Bhattacharyya, Hrishikesh Bhaumik, Sourav De, Goran Klepac, Handbook of Research on Intelligent Analysis of Multimedia information (Hardcover),Chapter 12: Time-Slot Based Intelligent Music Recommender in Indian Music, ISBN13: 9781522504986, ISBN10: 1522504982, Publisher: IGI Global, USA, 2016.
- 19. References Contd…. Sudipta Chakrabarty, Samarjit Roy, debashis De, “A Foremost Survey on State-Of-The-Art Computational Music Research”, In Proceedings of the Recent Trends in Computations and Mathematical Analysis in Engineering and Sciences”, International Science Congress Association, pp. 16-25, 2015. Sudipta Chakrabarty, Samarjit Roy, debashis De, “Behavioural Modelling of Ragas of indian Classical Music using Unified Modelling Language”, In Proceedings of the 2nd International Conference on Perception and Machine Intelligence, ACM, pp. 151-160, 2015. Sudipta Chakrabarty, Samarjit Roy, Debashis De, “Automatic Raga Recognition using Fundamental Frequency Range of Extracted Musical Notes” Accepted In: Eight International MultiConference on Image and Signal Processing (ICISP 2014), Elsevier, pp. 337-345, 2014. Sudipta Chakrabarty, Samarjit Roy, Debashis De, "Pervasive Diary in Music Rhythm Education: A Context-Aware Learning Tool Using Genetic Algorithm." Advanced Computing, Networking and Informatics-Volume 1. Springer International Publishing, pp. 669-677, 2014.
- 20. References Contd…. Samarjit Roy, Sudipta Chakrabarty, Pradipta Bhakta, Debashis De, “Modelling High Performing Music Computing using Petri Nets,” Accepted In: International Conference on Control, Instrumentation, Energy and Communication (CIEC), IEEE, pp. 678-682, 2014. Samarjit Roy, Sudipta Chakrabarty, Debashis De, "A Framework of Musical Pattern Recognition Using Petri Nets." In the Proceedings of Emerging Trends in Computing and Communication. Springer India, pp. 245-252, 2014. Sudipta Chakrabarty, Debashis De, “Quality Measure Model of Music Rhythm using Genetic Algorithm”, In Proceedings of International Conference on Radar, Communication and Computing (ICRCC), IEEE, pp. 125-130, 2012.