Simple and complex activity cooperative identification method based on deep multitask learning

• Main Authors: PENG LIANGYING, CHEN LING
• Format: Patent
• Language: Chinese; English
• Published: 23.11.2018
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a simple and complex activity cooperative identification method based on deep multi-task learning. The invention comprises the following steps of: 1) performing time window division on original activity data to obtain simple and complex activity samples, wherein a complex activity sample consists of multiple simple activity samples; 2) extracting a simple activity characteristic through the CNN network, and establishing a simple activity classifier; 3) extracting complex activity timing characteristics through LSTM network, and establishing a complex activity classifier; 4) the two classification tasks sharing a CNN layer and a simple activity characteristic layer, and cooperative training a simple and complex activity classifier by a shared structure. 5) identifying the probability of activity and the complex activities of activity data to be detected by the trained simple and complex activity classifiers. The simple and complex activity cooperative identification method based on deep