多媒體處理核心技術及產業化, 嵌入式智慧多媒體資訊處理平臺
GHP/040/09CZ
20100301 - 20111231
4855
趙盛章博士
The proposal will develop core multimedia technologies for machine processing to achieve better content understanding and rendering.
• Design and development of high performance and reliable technology IP:
1. Embedded intelligent multimedia information processing platform, which supports 4-channels D1 or single channel 1080i H.264 video compression, and video scene analysis simultaneously in a single DSP chip.
2. Visual scene analysis algorithm, such as background estimation, object segmentation, object tracking using Feature Density Approximation(FDA), behavior analysis, fusion of video compression and scene analysis, which leverage the motion vector information in video compression to reduce the computation overhead in object tracking dramatically.
3. Enhanced impulsive sound detection algorithm, and new word spotting algorithm for 20 sensitive keywords.
4. Fusion of sound/video scene analysis
5. Automatic detection and recognition of intrusion or violence scenes in real time. Other scene matching with new scene template defined by customers are also included.
6. Audio & video de-noising algorithm
7. Object extraction algorithm for speaker and virtual background mixing
8. Integrated sound/video de-noising and high definition video compression technology .
9. Embedded intelligent multimedia information processing platform which can be used in intelligent surveillance and multimedia communication area, and achieve 30% cost down compare with similar market product.
• Open and flexible software module design, ready for integration with existing traditional application systems such as surveillance systems, IPTV systems, digital film archiving systems.1. Object tracking technology optimization
2. Video De-Noising technology, based on the de-noising and super resolution algorithm to develop optimized video de-noising on decoder side, trans-coding and multiple-view in server side, to improve the video quality under the mix-rate video network.
3. Net-True virtual reality development, it is based on the object segmentation algorithm developed by ASTRI, to implement conference Net-True technology, virtual single round table conference, and try to become the world leader in this domain.
4, Carrier level surveillance platform development, to be deployed in all over the country.
5. Wireless video surveillance, support multiple standard such as TD-SCDMA, WCDMA,CDMA 2000 and Wi-Max
6. Advanced streaming and storage technology. To achieve mass volume video live and storage via distributed architecture and traffic balance.
7. Carrier level intelligent surveillance server platform, which can support automatically detect and recognize intrusion or violence scenes in real time, set alarm, and request for immediate processing, Analysis the signals from group of cameras deployed in one area, so as to achieve tracking the suspect via camera network, and its commercialization.
深圳研發中心，寳利通通信技術（北京）有限公司 [贊助機構]
香港大學
中興通訊股份有限公司 [贊助機構]
本專案計畫研發智慧的多媒體資訊處理的核心技術平臺,以達致智慧的資訊創建, 針對緊急事件檢測的場景分析,表達和增強。 現存的基於人工監視的監控系統，依賴人工發現緊急事件。建議的智慧技術平臺，將能自動監測緊急事件，減少人工幹預程度，提高警方回應速度，改善治安，有利於創建和諧社會。 核心技術包括嵌入式多媒體資訊處理硬體平臺以及提供智慧監控和多媒體通信所需的演算法，包括高清晰度視頻H.264壓縮，降噪，智慧視音頻場景處理以及敏感辭彙識別的融合分析，行為分析，視頻增強技術。將同本地大學在核心演算法研究上合作。 隨著多媒體資訊工業，特別是智慧監控和多媒體工業在大中華地區的快速發展，本專案將產生巨大商機。