Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM68899
Volume:
10 |
11 |
Nov 2026
Pages:
1-10
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.

Department of Electronics and Communication
kalpataru institute of technology

Department of Electronics and Communication
kalpataru institute of technology

Department of Electronics and Communication
kalpataru institute of technology

Department of Electronics and Communication
kalpataru institute of technology

Department of Electronics and Communication
kalpataru institute of technology
Voice-based interfaces provide a convenient method of interacting with embedded systems. However, conventional cloud-connected voice assistants depend on network availability and may transmit sensitive voice data to remote servers. This paper presents the design of a secure offline voice authentication and command system using Tiny Machine Learning (TinyML) for edge devices. The proposed prototype uses an ESP32-family mi-crocontroller and an INMP441 I2S digital microphone to acquire voice input. The captured audio is processed locally, and voice features are extracted using Mel-Frequency Cepstral Coefficients (MFCC). A machine-learning model developed using the Edge Impulse workflow is intended to distinguish an authorized voice from an unauthorized voice. Based on the model output and the configured decision logic, the controller permits or denies the requested operation. For the access-control demonstration, a relay module interfaces with a 12 V solenoid door lock, while LEDs provide status indication. The principal objective is to perform authentication and command processing on the edge device without requiring an Internet connection during operation. This paper describes the system architecture, hardware and software design, audio processing workflow, TinyML model-development methodology, and functional evaluation procedure. Quantitative accuracy, false-acceptance rate, and false-rejection rate must be reported using the final independent test results obtained from the implemented model.
KEYWORDS
This paper presented the design of a Secure Offline Voice Authentication and Command System using TinyML for edge devices. The proposed prototype integrates an ESP32-family microcontroller with an INMP441 digital microphone to ac-quire voice input and perform local processing. An MFCC-based audio feature workflow and an Edge Impulse model-development process are used to develop a compact classifier for the configured authorized and unauthorized classes. The controller evaluates the model output and uses the authentication decision to control a relay connected to a 12 V solenoid door lock. LED indicators provide local status information. The architecture is designed so that audio acquisi-tion, inference, and access-control decisions can be performed without Internet connectivity during normal operation. The project team reports 150 authorized and 150 unautho-rized voice recordings, an overall model accuracy of 85%, and a fully working hardware prototype. Further independent testing across additional speakers and acoustic conditions can strengthen the evaluation. The system provides a practical plat-form for studying offline TinyML-based voice authentication on resource-constrained embedded devices.
[1] S. B. Davis and P. Mermelstein, “Comparison of parametric represen-tations for monosyllabic word recognition in continuously spoken sen-tences,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 28, no. 4, pp. 357–366, Aug. 1980. [2] D. A. Reynolds, T. F. Quatieri, and R. B. Dunn, “Speaker verification using adapted Gaussian mixture models,” Digital Signal Processing, vol. 10, no. 1–3, pp. 19–41, 2000. [3] P. Warden and D. Situnayake, TinyML: Machine Learning with Tensor-Flow Lite on Arduino and Ultra-Low-Power Microcontrollers. O’Reilly Media, 2019. [4] C. Banbury et al., “MLPerf Tiny Benchmark,” in Proceedings of the Neural Information Processing Systems Track on Datasets and Bench-marks, vol. 1, 2021.
Publication Date
11/12/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
4
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...