As part of final year B.Tech project, collected and analyzed bird sound
samples from openly available sources(Xeno-canto, Macaulay library etc).
Proposed methods to classify real life bird sound recordings using audio
processing and supervised machine learning techniques, to identify characteristics
of birds such as size, habitat, types of call and species.
Implemented audio segmentation based on previously published data set
on soccer videos which used MFCC features and Deep Belief Networks. Alternate
feature set such as pitch, energy & spectral flux can yield up to 88
% accuracy in classification. Developed machine learning models in Weka
& R, used Python scripts & Matlab MIRtoolbox for feature extraction.
Compares several supervised machine learning models which categorize Twitter
users into six interest categories Politics, Entertainment, Entrepreneurship,
Journalism, Science & Technology and Health care; based on tweet, user
and timeseries features.
We study the performance of DANE protocol at client side and also present
a tool for deploying and administrating DANE with BIND servers in a local
network.
