2D Camera-Based Air-Writing Recognition Using Hand Pose Estimation and Hybrid Deep Learning Model

Air-writing is a modern human–computer interaction technology that allows participants to write words or letters with finger hand movements in free space simple and intuitive manner. recognition particular case of gesture which gestures can be matched characters digits the air. Air-written show extensive variations depending on various writing styles their speed articulation, presents quite difficult task for effective character recognition. In order solve these difficulties, this current work proposes an air-writing system using web camera. The proposed consists two parts: alphabetic digit assess our system, datasets were used: dataset numeric dataset. We collected samples from 17 asked each participant (A Z) (0 9) about 5–10 times. At same time, we recorded position fingertips MediaPipe. As result, 3166 1212 First, preprocessed then created datasets: image data padding sequential data. fed into convolution neural networks (CNN) model, whereas bidirectional long short-term memory (BiLSTM). After that, combined models trained again 5-fold cross-validation increase accuracy. work, model referred as hybrid deep learning model. Finally, experimental results showed achieved alphabet accuracy 99.3% 99.5%. also validated another publicly available 6DMG Our provided better compared existing system.