Mentor: Prof. Preeti Rao
This is a currently on-going research project undertaken by me in summers.
The problem I have to solve is simple, I need to identify the bird species from their sounds. I have used a machine learning approach to find a solution of theis problem.
The basic fronts which I needed to work upon were (1) I needed to identify the distinguishing features from bird sounds and extract them form the training data autonomously (2) Segmentation of the usable syllabe from the input recording given. (3) Apply a machine learning algorithm to classify the test data into its appropriate training data cluster, with a certain measure of accuracy.
What I have accomplished till now is a pretty good system of accurately classifying bird specied based on their sound. Furthermore, this process is completely automated apart from the syllable segmentation part which I am currently working on. The future step would be to detect and correct outliers fromt the training dataset, improve the database and then finally convert this into a standalone tool using some priors like location of birds etc for better results.
Here is a part of the test results I got till now. It gives a great accuracy. Unfortunately I cannot make my research and my code avaliable online :P. But If viewers have any questions and are seeking any advise on this topic feel free to drop me a mail.
Will keep updation the status of the project as I make further progress