INTELLIGENT DISCIPLINE MAINTENANCE SYSTEM WITH CLIENT SERVER TOPOLOGY
Advisor:
Dr. Kashif Javed
Abstract:
Intelligent Discipline Maintenance System monitors non-disciplinary actions like door
slamming, whistling, screaming, smoking and garbage throwing. The system is divided into two
major parts. One is responsible for audio processing and other for video processing. The audio
detection and recognition system receives real-time audio from a microphone, use standardizing
procedures like Median Filters to normalize it, extract its features and form a feature matrix for
that sound sample. The extracted features are Zero Crossing Rate (ZCR), Short Time Energy
(STE) and Mel Frequency Cepstral Coefficients (MFCCs). In next stage, Principal Component
Analysis (PCA) is used to extract more descriptive features and reduce dimensionality of the
feature matrix. Finally, Support Vector Machine (SVM) using Gaussian kernel is trained and
later used to recognize sounds like door slamming, whistling and screaming. The video detection
and recognition system detects smoking and garbage throwing using image processing and
machine learning. It receives real-time video data and detects moving objects in the captured
video frame using background estimation method. Template matching is performed to detect
garbage or cigarette in the captured frame. In case of smoking event detection, persons face is
also detected using Haar Classifier. The smoke detection is carried out and the event is declared
as smoking or not smoking accordingly. When some non-disciplinary action is detected, the
system automatically reports the authority to take further actions.