| Summary: | Many people may not be aware of the adverse effects of noise pollution on their health which include hearing impairment, negative social behaviour, anxiety, sleep disturbances and intelligibility to understand speech. Machine learning (ML) is the concept of making the machine determines, classifies, and does operations without being explicitly programmed. It is used in many fields such as intelligent transportation system and autonomous driving. Research in audio recognition has traditionally focused on the domains of speech and music. Comparatively, little research was done towards recognizing non-speech environmental sounds. For this reason, this project aims to develop an ML based classifier of sounds originated from the environment and compares the sound levels with the recommended levels by international standards via a created Graphical User Interface (GUI). Noise Capture mobile application will be used to record four sources of environmental noise, that are from highway, railway, lawn mowers and birds. Then, Python programming will be used to simulate the classification model using Scikit-learn. The trained data entered Scikit-learn gathered from Support Vector Machine (SVM), K-Nearest Neighbours (KNN), Bootstrap Aggregation (Bagging) and Random Forest (RF) classifiers, as well as Artificial Neural Network (ANN) algorithm from Keras and TensorFlow libraries for comparative performances in the accuracy test. In addition to ML, a noise pollution survey is conducted to provide qualitative analysis of community perceptions. The findings of ML are presented in terms of confusion matrix, accuracy, precision, recall and F1 score. The results show that the noise classification accuracy for all models exceeded 95%. The best ML models are RF and ANN due to its high accuracy and the least computational time. The findings of survey are also presented, which indicates that there is no correlation between gender, age, location with knowledge of noise pollution and the effect of noise on people. People are bothered by noise regardless of their age and gender.
|
|---|