Projects

Master's Thesis:

Detection of Taan Sections in Khayal Vocal Concerts [report]

The work is concerned with the automatic segmentation of specific musical sections from the audio of Khayal vocal music concerts in Hindustani classical music. The taan section, has a distinct melodic character across concert, irrespective of the tempo. Our goal is to label the taan sections using acoustic features that capture the melodic style. The proposed system does binary classifi.cation of frames into taan and non-taan classes.

The posterior probability vectors obtained in the course of the statistical classifi.cation are used in grouping stage. A higher time-scale smoothing is achieved by using change detection methods. The grouping stage uses heuristics derived from a study of musicians' annotations of taan episodes on a concert data subset. We evaluate the system in two stages: by its frame level classifi.cation accuracy, and by reporting the number of (over-, under-segmentation, true) detected, false positive and false negative taan episodes after the grouping stage. We compare the results of our proposed method with an unsupervised segmentation method showing that the proposed method achieves superior results over a database of 96 concerts in terms of the giving fewer false positives.

The algorithm is expected to be integrated into the browser: http://dunya.compmusic.upf.edu/

[Back To Samples of Work]

Course Projects during Masters:

Improving Singing Voice Detection for Indian Classical Music using Machine Learning Techniques [report]

Proposed improvements to the current state of the art method and validated the need for genre specific features by comparing selected features with baseline features. Applied select supervised and unsupervised machine learning algorithms with k-means achieving 87% accuracy for Carnatic genre and 82% for Hindustani genre

Musical Instrument Synthesis using DSP Processor [report]

Implemented successfully the methods like additive synthesis, frequency modulation, Karplus-Strong algorithm for synthesis of musical sounds on MATLAB as well as in hardware using TMS320C6713 kit.Produced melody using sound of clarinet, flute, violin, cello, guitar and bells

Fetched highest marks among 30 project groups with bonus marks for hardware implementation

Optical Sheet Music Recognition [report]

Applied image processing methodologies like binarization, untilting, dilation, erosion, etc for automatic detection of the notes present in the sheet music. Tested the algorithm on clean images to achieve approximately 98% accuracy for note symbol detection but needs considerable improvement for noisy images

Memory Card based Wav Player [report]

Designed an audio player using Atmega32 microcontroller with memory card interface. Achieved continuous playing of audio by parallel running of the tasks, of storing the contents and playing the song. Internal PWM used in fast PWM generation mode for D/A conversion for playing the songs

Object Recognition using Wavelets [report]

Applied wavelet decomposition to Invariant Fourier-Wavelet and Radon transformed descriptors; these two algorithms were used to calculate features for 135 images of butterfly, pyramid and yin-yang. Achieved an average 80% classification accuracy using known and unknown images in database using wavelets like Daubechies-4 and Haar level 4

Frequency to Voltage Converter Circuit [report]

This project dealt with the design and development of frequency to voltage converter circuit with low ripple and signal duty cycle independence.

Comparison of different configurations for ripple reduction without much compromise on settling time, viz using (i) only a low pass filter, (ii) a sample and hold circuit followed by a low pass filter, and (ii) a cascade of sample and hold circuits, was done.

Best results were observed with single S/H circuit followed by Sallen-Key LPF with large frequency range from 200Hz-70 KHz and sensitivity gain as 1mV/100Hz.

[Back To Samples of Work]

Projects during under-graduation:

Guitar Tuner

Tuned the guitar approximately by comparing the input frequency with stored known value using microcontroller (Atmega 8535) and analog circuit. The aim was to help amateurs to tune their guitar. Each of the six strings of a guitar has a unique fundamental frequency depending on the scale to be tuned to. The project combined analog filter design, analog amplifier design, and an algorithm for the microcontroller. The custom designed pre-amplifier, filter and Schmitt Trigger were successfully interfaced with the microcontroller.

Discreet time control technique for Induction Heating System

A novel discrete time control technique for Induction heating system is discussed. Discrete time control system is modular and flexible compared to its continuous counterpart. This control system holds major advantages of both phase shift control and pulse density modulation. Texas Instrument's Delfino Series processor was used for this application.

Rapid Prototyping of Electrical Loads and Machines on an Embedded processor

Mathematical model for variable speed windmill was developed and implemented on embedded platform for hardware in loop simulations.

Control algorithms were implemented and verified using RTOS based frame work

Pico-hydro power generation system

5 kW hydro power generation system was designed with intelligent power delivery system.This project was funded by Department of Science and Technology, Government of India. Link and demo at the link

[Back To Samples of Work]