A fast and accurate motion descriptor for human action recognition applications

Abstract

Abstract-With the availability of the recent human skeleton extraction algorithm introduced by Shotton et al. [1], an interest for skeleton-based action recognition methods has been renewed. Despite the importance of the low-latency aspect in applications, it can be noted that the majority of recent approaches has not been evaluated in terms of computational cost. In this paper, a novel fast and accurate human action descriptor named Kinematic Spline Curves (KSC) is introduced. This descriptor is built by interpolating the kinematics of joints (position, velocity and acceleration). To overcome the anthropometric and the execution rate variabilities, we respectively propose the use of a skeleton normalization and a temporal normalization. For this purpose, a new temporal normalization method based on the Normalized Accumulated kinetic Energy (NAE) of the human skeleton is suggested. Finally, the classification step is performed using a linear Support Vector Machine (SVM). Experimental results on challenging benchmarks show the efficiency of our approach in terms of recognition accuracy and computational latency.

An extension of kernel learning methods using a modified Log-Euclidean distance for fast and accurate skeleton-based Human Action Recognition

“…This leads to a system able to take a faster decision. c) A more complete description of the action: As shown in (Zanfir et al, 2013;Ghorbel et al, 2016), the inclusion of kinematic values further to the position such as the velocity and the acceleration of joints boosts the results.…”

“…Since skeleton representation has been very often used in bio-mechanic studies (Johansson, 1973), many papers have based their work on kinematic entities such as position, velocity and acceleration of joints. These values are computed thanks to the joint position information (Zanfir et al, 2013;Ghorbel et al, 2016). Zanfir et al (2013) proposed to concatenate these features and to weight each term by an empirical value.…”

“…The velocity and the acceleration at an instant t k are also calculated as in (Zanfir et al, 2013;Ghorbel et al, 2016).…”

Kinematic Spline Curves: A temporal invariant descriptor for fast action recognition

“…Based on our previous work [17], the paper presents a complete action recognition This article follows this scheme: in Section 2, an overview of related methods is given. Section 3 presents the concept of cubic spline interpolation and Section 4 presents the methodology used to build our descriptor.…”

Features and Classification Schemes for View-Invariant and Real-Time Human Action Recognition

“…Accuracy (% ) Seidenari et al [15] 82.00 Anirudh et al [31] 89.67 Devanne et al [8] 87.04 Cippitelli et al [19] 76.10 Vemulapalli et al [7] 90.88 Ours 90.32 [32] 87.90 Slama et al [10] 88.50 Xia et al [14] 90.92 Vemulapalli et al [7] 97.08 Ours 91.10 [7] presents a better accuracy, this method remains very time consuming to compute its decision. In [33], the computational time of [7] has been evaluated on MSRAction3D dataset. They reported that the approach takes an average of 17.61s to compute a descriptor (mean time on a sequence), knowing that MSRAction3D contains segmented videos composed of 12 to 54 frames.…”