Serverless parallel video combiner with dominant speaker detection for ultra–high definition multipoint video communication systems

The unprecedented shift towards user ubiquity in the 21st century coupled with rapid advancements in computing and network infrastructures have significantly increased the mass adoption of multipoint video communication among consumers, governments and corporations globally. This rapid adoption has...

全面介紹

Saved in:
書目詳細資料
主要作者: Baskaran, Vishnu Monn
格式: Thesis
出版: 2015
主題:
標簽: 添加標簽
沒有標簽, 成為第一個標記此記錄!
實物特徵
總結:The unprecedented shift towards user ubiquity in the 21st century coupled with rapid advancements in computing and network infrastructures have significantly increased the mass adoption of multipoint video communication among consumers, governments and corporations globally. This rapid adoption has brought upon new technical challenges. The first technical challenge focuses on the inefficacies of the conventional centralised based video combiner architecture in terms of its scalability, computational efficiency and image quality. The second technical challenge pertains to the latency in stitching high resolution video frames for ultra-high definition (UHD) display systems in real-time. The third technical challenge emphasises on the variability of speech characteristics amongst conference participants. This variability gives rise to transient speech patterns that result in misclassification of a dominant speaker. This thesis proposes original solutions for the aforementioned challenges to further enhance the performance of a multipoint video communication system.