Design of a Virtual Reality-Supported Immersive English Learning Environment and Interaction Behavior Analysis

Abstract

To address limitations in current virtual reality (VR)-based English learning environments— specifically the insufficient accuracy of real-time speech recognition, restricted interactive feedback, and the lack of advanced behavioral analysis capabilities—a comprehensive solution integrating advanced speech recognition techniques with immersive system design was developed. The study is organized into two major components. First, an end-to-end speech recognition model tailored for immersive VR environments was designed and implemented. The model integrates a low-rank feed-forward module with a probabilistic sparse self-attention mechanism, substantially reducing computational complexity while maintaining real-time processing performance on consumer-grade VR devices. Furthermore, a dual-path decoder architecture combining Connectionist Temporal Classification (CTC) with an attention-based mechanism was employed. This design simultaneously satisfies the low-latency requirements of streaming recognition and leverages rescoring in non-streaming modes to achieve higher accuracy, thereby improving robustness and precision in recognizing continuous, natural English speech. Second, building upon this recognition model, a fully integrated VR-based English learning system was designed. The system generates contextualized language tasks, supports multimodal human-computer interaction, and comprehensively records both speech and behavioral data throughout the learning process. This approach not only provides learners with a highly immersive and interactive environment featuring immediate pronunciation feedback and personalized training but also establishes a robust data foundation and technical platform for in-depth analysis of interaction behaviors and mechanisms of language acquisition. These contributions highlight significant theoretical value and practical potential for advancing intelligent education.