With more multi-modal data available for visual classification tasks,human action recognition has become an increasingly attractive topic.However,one of the main challenges is to effectively extract complementary feat...With more multi-modal data available for visual classification tasks,human action recognition has become an increasingly attractive topic.However,one of the main challenges is to effectively extract complementary features from different modalities for action recognition.In this work,a novel multimodal supervised learning framework based on convolution neural networks(Conv Nets)is proposed to facilitate extracting the compensation features from different modalities for human action recognition.Built on information aggregation mechanism and deep Conv Nets,our recognition framework represents spatial-temporal information from the base modalities by a designed frame difference aggregation spatial-temporal module(FDA-STM),that the networks bridges information from skeleton data through a multimodal supervised compensation block(SCB)to supervise the extraction of compensation features.We evaluate the proposed recognition framework on three human action datasets,including NTU RGB+D 60,NTU RGB+D 120,and PKU-MMD.The results demonstrate that our model with FDA-STM and SCB achieves the state-of-the-art recognition performance on three benchmark datasets.展开更多
Moving object detection in video surveillance is an important step. This paper addresses an automatic object detection algorithm based on spatio-temporal compensation for video surveillance. Temporal difference of the...Moving object detection in video surveillance is an important step. This paper addresses an automatic object detection algorithm based on spatio-temporal compensation for video surveillance. Temporal difference of the pairs of two frames with a k-frame distance is utilized to obtain coarse object masks. Usually, object regions in these coarse masks have discontinuous boundaries and some holes. Region growing with the distance constraint is proposed to compensate these coarse object regions in spatial domain, followed by filling holes. The added distance constraint can prevent object regions from growing infinitely. The proposed filling holes method is simple and effective. To solve the temporarily stopping problem of moving objects, temporal compensation is proposed to compensate the object mask by utilizing temporal coherence of moving objects in temporal domain. The proposed detection algorithm can extract moving objects as completely as possible. Experimental results have successfully demonstrated the validity of the proposed algorithm.展开更多
Analyzed the relation between time delay difference and time delay estimation errors, based on the principles of three-point locating, a reformed threshold method for time delay difference estimation of impulse signal...Analyzed the relation between time delay difference and time delay estimation errors, based on the principles of three-point locating, a reformed threshold method for time delay difference estimation of impulse signals, called as amendment estimation for short, is developed by introducing channel equalization technique to its conventional version, named as direct estimation in this paper, to improve the estimation stability. After inherent relationship between time delay and phase shift of signals is analyzed, an integer period error compensation method utilized the diversities of both contribution share and contribution mode of concerned estimates is proposed under the condition of high precision phase lag estimation. Finally, a cooperative multi-threshold estimation method composed of amendment and direct estimations to process impulse signals with three thresholds is established. In sea trials data tests of passive locating, this method improves the estimation precision of time delay difference efficiently. The experiments verify the theoretical predictions.展开更多
基金This work was supported by the Natural Science Foundation of Guangdong Province(Grant Nos.2022A1515140119 and 2023A1515011307)the National Key Laboratory of Air-based Information Perception and Fusion and the Aeronautic Science Foundation of China(Grant No.20220001068001)+1 种基金Dongguan Science and Technology Special Commissioner Project(Grant No.20221800500362)the National Natural Science Foundation of China(Grant Nos.62376261,61972090,and U21A20487).
文摘With more multi-modal data available for visual classification tasks,human action recognition has become an increasingly attractive topic.However,one of the main challenges is to effectively extract complementary features from different modalities for action recognition.In this work,a novel multimodal supervised learning framework based on convolution neural networks(Conv Nets)is proposed to facilitate extracting the compensation features from different modalities for human action recognition.Built on information aggregation mechanism and deep Conv Nets,our recognition framework represents spatial-temporal information from the base modalities by a designed frame difference aggregation spatial-temporal module(FDA-STM),that the networks bridges information from skeleton data through a multimodal supervised compensation block(SCB)to supervise the extraction of compensation features.We evaluate the proposed recognition framework on three human action datasets,including NTU RGB+D 60,NTU RGB+D 120,and PKU-MMD.The results demonstrate that our model with FDA-STM and SCB achieves the state-of-the-art recognition performance on three benchmark datasets.
基金National Natural Science Foundation of China (No.60502034)
文摘Moving object detection in video surveillance is an important step. This paper addresses an automatic object detection algorithm based on spatio-temporal compensation for video surveillance. Temporal difference of the pairs of two frames with a k-frame distance is utilized to obtain coarse object masks. Usually, object regions in these coarse masks have discontinuous boundaries and some holes. Region growing with the distance constraint is proposed to compensate these coarse object regions in spatial domain, followed by filling holes. The added distance constraint can prevent object regions from growing infinitely. The proposed filling holes method is simple and effective. To solve the temporarily stopping problem of moving objects, temporal compensation is proposed to compensate the object mask by utilizing temporal coherence of moving objects in temporal domain. The proposed detection algorithm can extract moving objects as completely as possible. Experimental results have successfully demonstrated the validity of the proposed algorithm.
文摘Analyzed the relation between time delay difference and time delay estimation errors, based on the principles of three-point locating, a reformed threshold method for time delay difference estimation of impulse signals, called as amendment estimation for short, is developed by introducing channel equalization technique to its conventional version, named as direct estimation in this paper, to improve the estimation stability. After inherent relationship between time delay and phase shift of signals is analyzed, an integer period error compensation method utilized the diversities of both contribution share and contribution mode of concerned estimates is proposed under the condition of high precision phase lag estimation. Finally, a cooperative multi-threshold estimation method composed of amendment and direct estimations to process impulse signals with three thresholds is established. In sea trials data tests of passive locating, this method improves the estimation precision of time delay difference efficiently. The experiments verify the theoretical predictions.
