Abstract: Adaptive bitrate (ABR) streaming dynamically adjusts video bitrates according to the network conditions to improve the users' quality of experience (QoE). However, existing ABR algorithms do not adequately consider the relationship between video content features and perceptual video quality, which limits their performance. To solve this problem, a video-content-aware adaptive bitrate algorithm is proposed. Subjective video quality assessment is employed to quantify the impact of different content features on perceptual quality scores. A deep reinforcement learning (DRL) model is then used to capture the relationship between spatiotemporal video-content features and perceptual video variations. In addition, distributed parallel training is adopted to accelerate model convergence, while principal component analysis (PCA) is applied to reduce feature dimensionality and improve training efficiency. Experimental results demonstrate that the proposed algorithm consistently improves QoE by approximately 3% ~ 5% under both stable and fluctuating network conditions with varying video chunk sizes.