Ahsan Shabbir ¹ Dr. Abdul Haleem Butt ¹ Taha Khan ² Lorenzo Chiari ³ Ahmad Almadhor ⁴ Vincent Karovič ^5*

• ¹ Air University, Islamabad, Pakistan
• ² Detectivio AB, Goteborg, Sweden
• ³ Department of Electrical, Electronic and Information Engineering, School of Engineering, University of Bologna, Bologna, Emilia-Romagna, Italy
• ⁴ Department of Computer Science, College of Computer and Information Sciences, Jouf University, Sakaka, Al Jawf, Saudi Arabia
• ⁵ Department of Information Management and Business Systems, Comenius University, Bratislava, Slovakia

Ambient acoustic detection and localization are processes that are performed to detect the event and its origin from acoustic data. The primary objective of this investigation was to establish a comprehensive framework for the classification of activities in the home environment in order to detect emergency events and transmit emergency signals. Furthermore, localization facilitates the detection of the acoustic event's location, thereby enhancing the overall effectiveness of emergency services, situation awareness, and timely response times. The acoustic data was collected in the home using six microphones that were strategically positioned. A bedroom, kitchen, restroom, and corridor comprised the home. A total of 512 audio samples were gathered from eleven activities. The background noise that was present in the acoustic signals was eliminated using the filtering technique. The study concentrates on the development of efficient frameworks; therefore, we extracted state-of-the-art features from the time domain, frequency domain, time-frequency domain, and cepstral domain. The study achieved successful detection and localization results, with random forest and linear discriminant analysis classifiers achieving an Accuracy of 95% and 87% for event detection, respectively. Conventional algorithm estimation signal parameters through Rational-In-variance Techniques (ESPRIT) were implemented for sound source localization. We attained significant results, with an error rate of 3.61, an Mse of 14.98, and an Rmse of 3.87. In the end, detection and localization models were combined for emergency activity detection and transmitting emergency notifications via electronic mail. The results endorsed that the proposed methodology attained a major milestone in developing a real-time emergency alert system for the domestic environment.