ويژگي هاي نهفته در فضاي پيچش زماني پويا براي بازشناسي كنش با استفاده از داده هاي سن سور كينكت

Embedded Feature Representation in Dynamic Time Warping Space for 3D Action Recognition Using Kinect Depth Sensor

ﻫﺪف از اﯾﻦ ﻣﻘﺎﻟﻪ، ﺑﺎزﺷﻨﺎﺳﯽ ﮐﻨﺶ اﻓﺮاد ﺑﺎ اﺳﺘﻔﺎده از اﻃﻼﻋﺎت ﺳﺮيﻫﺎي زﻣﺎﻧﯽ اﺳﺘﺨﺮاج ﺷﺪه از دﻧﺒﺎﻟﻪﻫﺎي اﺳﻠﮑﺘﯽ ﺑﻪ ﻣﻨﻈﻮر اﺳﺘﻔﺎده در ﺳﺎﻣﺎﻧﻪﻫﺎي ﻣﺎﻧﯿﺘﻮرﯾﻨﮓ ﻓﻌﺎﻟﯿﺖﻫﺎي روزﻣﺮهي اﻧﺴﺎﻧﻬﺎ ﻣﯽﺑﺎﺷﺪ. ﺑﻪ اﯾﻦ ﻣﻨﻈﻮر، ﻫﺮ ﮐﻨﺶ ﺑﻪ ﺻﻮرت ﯾﮏ ﺳﺮي زﻣﺎﻧﯽ ﭼﻨﺪ ﺑﻌﺪي ﺑﯿﺎن ﺷﺪه و ﺳﭙﺲ ﺑﺎ اﺳﺘﻔﺎده از ﻣﻔﻬﻮم "ﺷﺒﻪ ﮐﺮﻧﻞ ﻣﺒﺘﻨﯽ ﺑﺮ ﻓﺎﺻﻠﻪي ﭘﯿﭽﺶ زﻣﺎﻧﯽ ﭘﻮﯾﺎ" ﺑﻪ ﯾﮏ ﻓﻀﺎي ﺑﺮداري ﻧﮕﺎﺷﺖ ﻣﯽﮔﺮدد. در اداﻣﻪ، ﺑﻪ ﻣﻨﻈﻮر اﺳﺘﻔﺎده از ﻧﺴﺒﺖ ﻫﻤﺒﺴﺘﮕﯽ-ﺗﻤﺎﯾﺰِ دﻧﺒﺎﻟﻪﻫﺎ در ﭘﺮوﺳﻪي ﺷﻨﺎﺳﺎﯾﯽ، اﯾﻦ ﻓﻀﺎي ﺑﺮداري ﺗﻮﺳﻂ روش ﻓﯿﺸﺮ ﺗﻨﻈﯿﻢ ﺷﻮﻧﺪه ﺑﻪ ﯾﮏ ﻓﻀﺎي ﺗﻤﺎﯾﺰي ﻧﮕﺎﺷﺖ ﺷﺪه و ﺗﺼﻤﯿﻢ ﮔﯿﺮي ﻧﻬﺎﯾﯽ در ﺧﺼﻮص ﻣﺤﺘﻮاي ﺣﺮﮐﺖ در ﻓﻀﺎي ﺣﺎﺻﻞ اﻧﺠﺎم ﻣﯽﭘﺬﯾﺮد. ﺑﺮ ﺧﻼف ﺳﺎﯾﺮ روش ﻫﺎي ﮐﺮﻧﻠﯽ ﻣﻮﺟﻮد، اﻟﮕﻮي ﻫﻤﺘﺮازي ﺣﺎﺻﻞ از ﭘﯿﭽﺶ زﻣﺎﻧﯽ، ﻣﻮﺟﺐ ﻣﯽﺷﻮد ﺗﺎ اﺛﺮ ﺷﯿﻔﺖ، و اﻧﻘﺒﺎض و اﻧﺒﺴﺎط ﻫﺎي زﻣﺎﻧﯽ دﻧﺒﺎﻟﻪﻫﺎ در ﻓﻀﺎي ﮐﺮﻧﻞ ﺑﻪ ﮐﻤﺘﺮﯾﻦ ﻣﯿﺰان ﻣﻤﮑﻦ ﮐﺎﻫﺶ ﯾﺎﺑﺪ. ﻫﻤﭽﻨﯿﻦ، روش ﻣﺎ ﭘﯿﭽﯿﺪﮔﯽﻫﺎي ﻣﺤﺎﺳﺒﺎﺗﯽ و ﻣﺤﺘﻮاﯾﯽ ﻣﻮﺟﻮد در اﺳﺘﺨﺮاج وﯾﮋﮔﯽ-ﻫﺎي اﺳﺘﺎﺗﯿﮏ و دﯾﻨﺎﻣﯿﮏِ دﻧﺒﺎﻟﻪﻫﺎي ﺣﺮﮐﺘﯽ را ﺣﺬف ﻧﻤﻮده و در ﻣﻘﺎﺑﻞ، آﻧﻬﺎ را در ﻗﺎﻟﺐ اﻟﮕﻮي ﻫﻤﺘﺮازي در ﻓﻀﺎي ﺑﺮداري ﮐﺮﻧﻞ ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﻣﯽدﻫﺪ. ﻧﺘﺎﯾﺞ ارزﯾﺎﺑﯽﻫﺎ ﺑﺮ روي ﺳﻪ ﭘﺎﯾﮕﺎه دادهي ﻣﻌﺮوف UTKinect ،TST و UCFKinect،ﻗﺎﺑﻞ رﻗﺎﺑﺖ ﺑﻮدن ﻋﻤﻠﮑﺮد روش اراﺋﻪ ﺷﺪه ﺑﺎ ﺑﺮﺗﺮﯾﻦ روﺷﻬﺎي ﻣﻮﺟﻮد در ﺑﺎزﺷﻨﺎﺳﯽ ﮐﻨﺸﻬﺎي اﻧﺴﺎﻧﯽ را ﻧﺸﺎن ﻣﯽدﻫﺪ.

This paper proposes a novel 3D action recognition technique which uses the skeletal information extracted from depth image sequences. First, each action is represented by a multidimensional time series where each dimension represents the position variation of one skeleton joint over time. The time series is then mapped into the kernel Hilbert space using a metric defined by Dynamic Time Warping distance. Afterwards, regularized Fisher strategy is used to remap the kernel space into a discriminative one. This incorporates the correlation-distinctiveness relationship of the sequences into the recognition process and also mitigates the curse of dimensionality effect in the kernel space. Unlike traditional kernel functions, the time warping used in the mapping strategy makes the kernel space robust to the temporal shift variations of the motion sequences. Moreover, our method eliminates the need for a complex design method for extracting the static and dynamic information of a motion sequence. A set of extensive experiments on three publically available databases; TST, UTKinect, and UCFKinect demonstrates the superiority of our method compared to a set of baseline algorithms.