Danger Detection for Women and Child Using Audio Classification and Deep Learning
Md Ashikuzzaman, Awal Ahmed Fime, Abdul Aziz, Tanvira Tasnima · 2021
Due to rising crime against women and children, it is essential to ensure security for them. Various modern techniques are being used nowadays for providing security. Some of them are sensor based devices, and others are mobile applications. But hardware based devices don’t work appropriately when sensors are disconnected from the body. Also, existing applications on mobile devices don’t work automatically. Victims need to make particular reactions like shaking the phone and pressing the SOS button, which might not always be possible. As audio classification is one of the most advanced applications of deep learning, we propose an idea to provide security for women and children applying audio classification. We have detected danger from the screaming sounds of the victim, which is detectable from a quite good distance. A few deep neural network models are used for audio classification, so various audio and reaction screaming can be considered a sign of danger. Additionally, we have created a new dataset for this purpose.