NTT Communications has developed an artificial-intelligence (AI) technology, called Time-series Deep Learning that analyzes time-series video data to identify specific human motions with high precision. In recent tests, NTT Com confirmed that Time-series Deep Learning can analyze video images taken with network cameras to distinguish between specific motions, such as people crouching, acting restlessly or moving things. Specific motions were identified with more than 80% accuracy. Existing technology mainly uses static images to analyze two-dimensional vertical and horizontal data.
Time-series Deep Learning detects motions more precisely using three-dimensional data by adding a time axis. It can also analyze IoT data as it changes over time, such as temperatures or voltages recorded with sensors. Conventional deep learning (machine learning) methods use one-frame images for convolutional bonding to extract features by learning data locations and points. Time-series Deep Learning convolutes pixels in multiple sequential frames. Possible applications include detection of specific motions or behavior for preventing crimes or accidents in factories, understanding purchasing behavior, analyzing sports perform.
