International Journal of Science, Engineering and TechnologyAuthors: Harsh Shah, Kedar Pinniboyina, Anand Makawana
Abstract: Voice assistant systems have become increasingly im- portant in modern computing environments, pro- viding hands-free interaction capabilities for users across different languages and contexts. Traditional voice assistants often require manual language se- lection and rely on proprietary APIs with limited customization options. This paper presents a novel multi-language voice assistant system that automati- cally detects user language (English, Hindi, Gujarati) and integrates multiple free web services for weather, search, and task management. The proposed sys- tem utilizes the Web Speech API for speech recog- nition and synthesis, implements a command pro- cessing engine with intent recognition, and provides real-time service integration with fallback mecha- nisms. Our system achieves 95% overall accuracy in voice recognition across three languages, with En- glish achieving 97%, Hindi 93%, and Gujarati 92% accuracy. The system responds to commands in an average of 1.2 seconds, significantly faster than tradi- tional voice assistants (3-5 seconds). Service reliabil- ity reaches 99.2% through intelligent fallback mech- anisms. The implementation demonstrates the effec- tiveness of browser-based voice processing combined with intelligent service orchestration for creating ac- cessible, multilingual voice interfaces.
DOI: https://doi.org/10.5281/zenodo.17092382
