EBi360: An Edge-Assisted Viewport Prediction Method for 360° Video Based on BiLSTM

360° video offers an immersive experience but also poses challenges in terms of substantial bandwidth requirements and latency sensitivity. To address these challenges, we propose an edge-assisted viewport prediction method for 360° video based on Bidirectional Long Short-Term Memory (BiLSTM) networks, named EBi360. This method leverages the bidirectional network structure to capture temporal context information and integrates object detection technology, combining deep learning with video content analysis to enhance viewport prediction accuracy. Moreover, we devise an adaptive bitrate allocation strategy that dynamically optimizes the distribution of bitrates across tiles in each chunk, thereby reducing bandwidth waste. Experimental results demonstrate that EBi360 outperforms other methods in terms of prediction accuracy and user quality of experience (QoE) metrics.