SecurePath: An AI-Powered Companion for Real-Time Personal Safety

Abstract

Personal safety is a growing concern, particularly for women and individuals traveling alone in urban environments. This paper presents SecurePath, an AI-powered voice-driven safety system designed to detect emotional states and trigger words in real time to enhance user security. The system integrates Speech Emotion Recognition (SER) and trigger-word detection using machine learning models based on MFCC feature extraction. A hybrid CNN–BiLSTM model with attention mechanism is employed for accurate emotion classification, while a CNN–LSTM model is used for efficient detection of emergency keywords.In addition to voice analysis, the system incorporates location-based unsafe zone detection to identify potentially risky environments. By combining emotional cues with contextual location data, SecurePath can automatically trigger alerts and notify trusted contacts when signs of distress or danger are detected. The models are implemented using TensorFlow and enhanced with data augmentation techniques to improve robustness and real-world performance.The proposed system demonstrates the effectiveness of AI-driven, context-aware safety solutions that operate without requiring manual intervention, offering a reliable approach to real-time personal security.

Introduction

SecurePath, an AI-powered voice-based personal safety system designed to improve security for individuals, especially women and solo travelers in urban areas. The system uses Speech Emotion Recognition (SER) and trigger-word detection to identify distress or emergency situations in real time.

It employs machine learning models such as a CNN–BiLSTM with attention mechanism for detecting emotions from speech and a CNN–LSTM model for recognizing emergency keywords. The system extracts speech features using MFCC (Mel Frequency Cepstral Coefficients) and uses TensorFlow for implementation, along with data augmentation to improve accuracy and robustness.

In addition to voice analysis, SecurePath includes location-based unsafe zone detection, combining emotional signals and geographic context to assess risk. When danger or distress is detected, the system automatically triggers alerts and notifies trusted contacts.

Conclusion

This paper presented SecurePath, an AI-powered personal safety system that integrates Speech Emotion Recognition (SER), Trigger Word Detection (TWD), and location-aware unsafe-zone detection for proactive safety assistance. Unlike conventional safety applications that rely on manual SOS activation, SecurePath continuously monitors user speech and environmental context to automatically identify distress situations.Experimental evaluation shows that the proposed SER model achieves 94.1% accuracy, while the optimized trigger word detection model reaches 96.4% accuracy using binary classification. The unsafe-zone detection module further supports the system by identifying risky locations with 88.5% accuracy and low response time. These results demonstrate that SecurePath can reliably detect unsafe situations and operate effectively for real-time personal safety applications. Future enhancements include training the models on more diverse speech datasets to improve robustness across accents and noisy environments, integrating additional contextual inputs such as wearable or motion data, and refining location-based risk analysis using dynamic crime data. Further optimization will enable efficient deployment on resource-constrained mobile devices.

References

[1] Ritu Chaudhary, Shreya Saraswat, Siddhant Chaturvedi, and Prof. Prajakta Naregalkar, “Speech Emotion Recognition Using Neural Network,” International Research Journal of Engineering and Technology (IRJET), vol. 7, no. 5, pp. 4050–4053, May 2020 [2] Seme Sarker, Khadija Akter, and Nursadul Mamun, “A Text Independent Speech Emotion Recognition Based on Convolutional Neural Network,” International Journal of Speech Technology, vol. 26, pp. 1–13, 2023. [3] Sheetal R. Waghmare and Prakash S. Nalbalwar, “Application of Probabilistic Neural Network for Speech Emotion Recognition,” International Journal of Innovative Science and Research Technology, vol. 9, no. 2, pp. 198–202, 2024. [4] Qianhe Ouyang, “Speech Emotion Detection Based on MFCC and CNN-LSTM Architecture,” International Journal of Innovative Technology and Exploring Engineering (IJITEE), vol. 13, no. 1, pp. 1–6, 2024. [5] Shreya Jena, Sneha Basak, Himanshi Agrawal, Bunny Saini, Shilpa Gite, Ketan Kotecha, and Sultan Alfarhood, “Developing a Negative Speech Emotion Recognition Model for Safety Systems Using Deep Learning,” Frontiers in Human Dynamics, 2025. [6] Eduard Franti, Ioan Ispas, Voichita Dragomir, Monica Dascalu, Elteto Zoltan, and Ioan Cristian Stoica, “Voice Based Emotion Recognition with Convolutional Neural Networks for Companion Robots,” International Conference on Automation, Quality and Testing, Robotics (AQTR), 2017. [7] Zhengwei Huang, Ming Dong, Qirong Mao, and Yongzhao Zhan, “Speech Emotion Recognition Using CNN,” ACM International Conference on Multimedia, pp. 801–804, 2014. [8] Ruhul Amin Khalil, Edward Jones, Mohammad Inayatullah Babar, et al., “Speech Emotion Recognition Using Deep Learning Techniques: A Review,” IEEE Access, vol. 7, pp. 117327 117345, 2019. [9] Siddique Latif, Rajib Rana, Sara Khalifa, Raja Jurdak, and Julien Epps, “Direct Modelling of Speech Emotion from Raw Speech,” arXiv preprint arXiv:1904.03833, 2019. [10] Chawki Barhoumi and Yassine BenAyed, “Real-Time Speech Emotion Recognition Using Deep Learning and Data Augmentation,” Artificial Intelligence Review, 2024. [11] Zheng Zeng and Liang Xiao, “Effective Combination of DenseNet and BiLSTM for Keyword Spotting,” 2018 IEEE 27th International Conference on Computer Communication and Networks (ICCCN), Hangzhou, China, pp. 1–8, 2018. [12] David Bonet, Guillermo Cámbara, Fernando López, Pablo Gómez, Carlos Segura, and Jordi Luque, “Speech Enhancement for Wake-Up-Word Detection in Voice Assistants,” ICASSP 2021, pp. 4567 4571, 2021. [13] Arnav Kundu, Mohammad Samragh, Minsik Cho, Priyanka Padmanabhan, and Devang Naik, “HEiMDaL: Highly Efficient Method for Detection and Localization of Wake-Words,” in ICASSP 2023 – IEEE International Conference on Acoustics, Speech and Signal Processing, 2023, pp. 1-5. [14] Rajath Kumar, Vaishnavi Yeruva, and Sriram Ganapathy, “On Convolutional LSTM Modeling for Joint Wake-Word Detection and Text Dependent Speaker Verification,” Proceedings of Interspeech, pp. 3313–3317, 2018. [15] Kenichi Kumatani, Sankaran Panchapagesan, Minhua Wu, Minjae Kim, Nikko Ström, Gautam Tiwari, and Arindam Mandal, “Direct Modeling of Raw Audio with DNNs for Wake Word Detection,” IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), 2017. [16] Zehetner, M. Hagmüller, and F. Pernkopf, “Wake-up-word spotting for mobile systems,” in 2014 22nd European Signal Processing Conference (EUSIPCO), Lisbon, Portugal, Sept. 1–5, 2014, pp. 1527–1531. [17] Yiming Wang, Hang Lv, Daniel Povey, Lei Xie, and Sanjeev Khudanpur, “Wake Word Detection with Alignment-Free Lattice-Free MMI,” Interspeech 2020, pp. 1234–1238, 2020. [18] Liliane Momeni, Triantafyllos Afouras, Themos Stafylakis, Samuel Albanie, and Andrew Zisserman, “Seeing Wake Words: Audio-Visual Keyword Spotting,” ICASSP 2020, pp. 5678–5682, 2020. [19] Yiming Wang, Hang Lv, Daniel Povey, Lei Xie, and Sanjeev Khudanpur, “Wake Word Detection with Streaming Transformers,” arXiv preprint arXiv:2103.01234, 2021. [20] Dr. Sridhar Mandapati, Sravya Pamidi, and Sriharitha Ambati, “A Mobile Based Women Safety Application (I Safe Apps),” International Journal of Computer Science and Mobile Computing (IJCSMC), vol. 4, no. 3, pp. 252–258, 2015. [21] Dr. R. Priya, Chiranchievi S, Santhosh M, Udhaya Parameshwaran M, and Tamizhbala S, “Safe Alert App,” International Journal of Scientific Research in Engineering and Management (IJSREM), vol. 9, no. 3, pp. 15–22, 2025. [22] Prof. Usha Kumari V, K Karthik, K Chetan, Chaithra K, and Bharath S, “SOS: Bridging the Gap Between Threat and Safety for Women Using Deep Learning Models,” International Journal of Engineering Research & Technology (IJERT), vol. 13, no. 2, pp. 56–63, 2024. [23] Sneha Dole and Dr. Santosh Jagtap, “WE SAFE App – A Women’s Safety Application Using AI,” International Conference on Smart Technologies and Systems for Next Generation Computing (ICSTSN), 2025. [24] https://www.kaggle.com/datasets/uwrfkaggler/ravdess-emotional-speech-audio last accessed on 16 September 2025.

Copyright

Copyright © 2026 Archie Patil, Pooja Narkar, Pradnya Lohar, Dr. Ashwini Save. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.