i -Honk: Self Honk Detection from Smartphone Sensing for Driver Personalization

Densification in traffic day by day leading to increase noise pollution and which may be the potential health issues on the local level
^1
1
https://www.sciencedirect.com/science/article/pii/S016041202300140X. One of the main causes of traffic congestion is the rapid growth of two-wheeler and four-wheeler particularly in emerging countries. We need driver profiles in order to take appropriate action based on honking in addition to good urban planning in order to control noise pollution. In this study, we use deep learning method to develop a smartphone-based self-honk detecting system based on two-wheeler. Here we have created two models, one for self-honk and ambient noise detection in controlled environment and another will classify self-honk, ambient honk and ambient noise in realtime environment. Here we get some useful results making the system into two models. In a controlled setting, Model 1 provides 98% training accuracy and 96 % testing accuracy in honk detection. In contrast, Model 2 has been trained in realtime using self-honk data, ambient noise, and ambient honk, and it provides 85 % training accuracy and 82 % testing accuracy.