基于MediaPipe与机器学习模型融合的手势识别框架

Gesture Recognition Framework Based on Fusion of MediaPipe and Machine Learning Model

文章提出了一种基于MediaPipe与机器学习模型融合的手势识别框架。首先,通过MediaPipe获取手部21个关键点;其次,使用数据集增强模块扩充数据集,根据角度信息和距离信息使用自动化特征构建模块生成特征并使用主成分分析进行降维;最后,使用机器学习模型在自主数据集进行训练和预测,通过验证发现支持向量机模型效果最好。

The article proposes a gesture recognition framework based on the fusion of MediaPipe and machine learning models.Firstly,21 key points of the hand are obtained by MediaPipe;secondly,the dataset is expanded using the dataset enhancement module,and features are generated based on angle information and distance information using the automated feature building module and dimensionality reduction using principal component analysis;finally,machine learning models are used to train and predict on the autonomous dataset,and the support vector machine model is found to work best through validation.