CAJ | 학술논문

目前基于手势识别的人机交互系统还没得到广泛应用，而且市面上大多数的应用都需要依赖昂贵的硬件平台。为了实现一个成本低廉且处理与识别速度高的动态手势识别控制系统，提出了一种基于查找多核高斯肤色哈希表和运动特征融合的快速识别算法，并用该算法在JZ4755嵌入式平台上构建了动态手势控制方案。该系统首先通过间隔降采样法减少处理的数据量，然后对采集的手势帧用查找高斯肤色哈希表和帧间异或运算提取运动特征区域，最后根据观察运动特征区域面积的突变特征确定输出手势指令来操作电视机播放。在摄像头输入视频帧分辨率为640*480的条件下，该系统的处理速度达到23帧/秒。实验结果表明，该人机交互系统具有方案实时性高、算法对硬件平台依赖低和操作人性化等特点。
목전기우수세식별적인궤교호계통환몰득도엄범응용，이차시면상대다수적응용도수요의뢰앙귀적경건평태。위료실현일개성본저렴차처리여식별속도고적동태수세식별공제계통，제출료일충기우사조다핵고사부색합희표화운동특정융합적쾌속식별산법，병용해산법재JZ4755감입식평태상구건료동태수세공제방안。해계통수선통과간격강채양법감소처리적수거량，연후대채집적수세정용사조고사부색합희표화정간이혹운산제취운동특정구역，최후근거관찰운동특정구역면적적돌변특정학정수출수세지령래조작전시궤파방。재섭상두수입시빈정분변솔위640*480적조건하，해계통적처리속도체도23정/초。실험결과표명，해인궤교호계통구유방안실시성고、산법대경건평태의뢰저화조작인성화등특점。
The human-computer interaction system based on gesture recognition is not used widely,most of the applications on the market depend on expensive hardware platform. In order to achieve a low cost and a high speed of dynamic gesture recognition control system, put forward a kind of fast recognition algorithm which based on searching multi-core Gaussian color hash table and integrating movement characteristics,then use the algorithm on JZ4755 embedded platform to build a dynamic gesture recognition control system. First,the sys-tem uses interval sampling method to reduce the amount of data. Then,splite the acquisition of gesture frames into exclusive area of mo-tion feature by searching Gaussian color hash table and inter-frame method. Last,according to observing mutation characteristics of move-ment area to output gesture indicative to operate the television broadcast. On the conditions which camera input frames size is 640*480, the processing speed of the system reach 23 FPS. The experimental results show that the human-computer interaction system has the fea-ture of high real-time,relying on hardware platform lowly,humanized operation and so on.