基于改进KNN算法的动态手势识别研究

Research on Gesture Recognition Based on Improved KNN Algorithm

手势识别是人机交互的一种方式,用于手势识别的传统K最邻近算法由于训练组数据量大影响了其识别效率,为此提出了一种新的手势特征提取方法,设计了一款基于改进K最邻近算法的手势识别俄罗斯方块游戏.该方法根据手势信号的特征量,只需记录特征量的符号作为训练组以及测试组来储存.实验表明,改进K最邻近算法在体感游戏中对手势识别的平均成功率较阈值判别法的手势识别成功率提高了10%左右.

Gesture recognition as a way of human-computer interaction.In view of the problem that the traditional K-Nearest Neighbor(KNN)algorithm in gesture recognition affected the recognition efficiency due to the cumbersome training group data,a new gesture feature extraction method was proposed,and a Tetris game of gesture recognition with improved KNN algorithm was designed.According to the characteristic quantity of the gesture signal,only the symbol of the recorded characteristic quantity can be stored as the training group and the test group.The experiment shows that the average success rate of gesture recognition by improved KNN algorithm in somatosensory games is about 10%higher than that by threshold discrimination.