<![CDATA[This research develops a voice recognition-based home door security systems using the BERT model and synonym data augmentation techniques to enhance the accuracy and responsive of the system. The objective of this research is to develop a system that can accurately and rapidly recognize voice commands and function effectively in a range of environments. This research employs the “bert-base-uncased” version of the BERT model, which has been trained on a dataset comprising 382 sentence variations. The synonym data augmentation technique is employed to extend the range of recognized commands. The system was subjected to a series of tests to assess its performance in terms of voice detection frequency, command categorization accuracy, response time, noise influence, and hardware integration. The results demonstrated that the BERT model achieved an accuracy of 99.29% and F1 score of 0.9929. The system exhibited an average response time of 0.12 seconds. However, the accuracy declined when the noise level exceeded 40 dB. The findings of this research indicate that the BERT model is highly effective in classifying voice commands in home security systems and has significant potential for real-time applications. The synonym data augmentation technique enhances the system’s capacity to process sentence variations. This research makes a significant contribution to the development of the dataset and optimization of noise handling for further improvement.]]>
Copyrights © 2025
