Mixed Signature: An Invariant Descriptor for 3D Motion Trajectory Perception and Recognition

Abstract

Motion trajectory contains plentiful motion information of moving objects, for example, human gestures and robot actions. Motion perception and recognition via trajectory are useful for characterizing them and a flexible descriptor of motion trajectory plays important role in motion analysis. However, in the existing tasks, trajectories were mostly used in raw data and effective descriptor is lacking. In this paper, we present a mixed invariant signature descriptor with global invariants for motion perception and recognition. The mixed signature is viewpoint invariant for local and global features. A reliable approximation of the mixed signature is proposed to reduce the noise in high-order derivatives. We use this descriptor for motion trajectory description and explore the motion perception with DTW algorithm for salient motion features. To achieve better accuracy, we modified the CDTW algorithm for trajectory matching in motion recognition. Furthermore, a controllable weight parameter is introduced to adjust the global features for tasks in different circumstances. The conducted experiments validated the proposed method.

Parsing 3D motion trajectory for gesture recognition

“…In recent years, several methods [8,9,[32][33][34][35] focus on using spatial shape invariants to represent 3D motion features and parameters. Rao et al [8] use curvature to calculate the instants and intervals of motion which are too simple to fully represent complex motions.…”

“…Rao et al [8] use curvature to calculate the instants and intervals of motion which are too simple to fully represent complex motions. Wu and Li [32] and Yang [33] use the differential invariants to represent the 3D motion trajectory directly without the 3D-2D projection, where most of the 3D information is preserved. After that, Shao and Li [35] extends the integral invariants [36][37][38] to the 3D space to represent 3D motion trajectories, which is more robust to high frequency noise than the differential invariants.…”

“…The signature descriptor [32] fails to maintain the invariance in this form of scale variation, which affects the calculation of the trajectory parameters in that descriptor. This problem is also indicated in [33] and is solved by adding global parameters as additional information into the descriptor for better accuracy, but that method is still not invariant to this variation.…”

Comparison of rigid body motion trajectory descriptors for motion representation and recognition

“…The motion trajectory of an object is used in various robotic applications such as motion recognition in humanrobot interaction [1], motion analysis and perception [2], and learning object manipulation tasks in robot programming by human demonstration [3]. The motion trajectory serves as an important tool for representing motion in addition to information about the object's properties, the environment in which the motion happens, or the pose of the manipulator holding the object.…”

View-Invariant Gesture Recognition Using Nonparametric Shape Descriptor

“…Some other works use global geometry features instead. Reference [13] added an additional global item on the basis of [12] to create a mixed descriptor and slightly enhanced the performance.…”