Blink talk: A machine learning-based method for women safety using EEG and eye blink signals

Abstract

Women's safety is currently thought to be a big issue in both urban and rural settings. A variety of smart gadgets and software were created to provide women with security. There are a lot of smart gadgets and applications on the market, but they don't offer a good answer and are too expensive. In this research, a novel machine learning-based Blink Talk method has been proposed for women safety. EEG based on some blink talk algorithms for eye blink detection. Initially, the EEG and eye blink signals are pre-processed using Discrete wavelet transform (DWT). The pre-processed signals are fed into Stacked Denoising autoencoder (SDAE) for extracting the features. In the next phase, the extracted features are used to classify the emotions of women through Multiclass Support vector machine (MSVM). The classification results are sad, happy, normal, and fear; finally, fear emotion is converted into text using GSM to the saved contacts and nearby police station and GPS scan the near radius surrounding people to send the alert and help request message for the help needed person. The experimental results reveal that the suggested approach provides high accuracy range of 98.04%. then the traditional machine learning techniques.

Keywords

• Women safety
• Machine learning
• Stacked denoising autoencoder (SDAE)
• Multiclass support vector machine (MSVM)
• EEG
• Discrete wavelet transform