Wei Ping Loh

Human motion classification using 2D stick-model matching regression coefficients

A 2D human motion model developed based on polynomial regression data fit.Human motion modeled in three segments: backbone, upper body and lower body.Tolerance rate considers the acceptable range for reliable motion estimations.Classification accuracy for developed estimation model is on par with the actual. Motion estimation methods have been proposed via different approaches, such as silhouette based, model based and image based estimations. However, these methods are highly dependent on the quality of motion data for optimal classification accuracy. Further, because of the complexity of existing algorithms for motion estimation, there are difficulties in interpretation. Hence, the contribution of this work is to model simple human motions for the purpose of recognizing different activity behavior patterns for classification analysis. The model is made up of three body component integrations - Backbone (BB), Upper Body (UB) and Lower Body (LB) - to form a simple 2D human stick figure. Two case studies involving a publicly available video of walking, running and jumping motions as well as experimental captures of Yoga motions are studied. Video motions are simplified into time-step image snapshots, which are later translated into a numeric 2D coordinate system. Initially, the human pelvis is considered the origin of the stick figure. The stick model was drawn by integrating the BB, UB and LB components based on the 2D body joint coordinates. The motion estimation model applies the concept of polynomial fitting to the coordinates data. Computations on the polynomial fitting coefficient deviations at sequential time steps were performed to evaluate the estimation tolerance. A summation of the precedent time-step coordinates with the average deviation metric is used iteratively to estimate the joint coordinates of the stick figure in the subsequent time step to develop the entire motion model. Finally, the developed motion estimation mathematical model was compared to the actual motion phases for classification efficiencies using the Bayes, Lazy, Function, Meta, Misc, Rules and Trees classifiers. Our findings revealed the feasibility of using 2D stick-model matching estimation for human motion classification analysis.

Data Treatment Effects on Classification Accuracies of Bipedal Running and Walking Motions

Many real-world data can be irrelevant, redundant, inconsistent, noisy or incomplete. To extract qualitative data for classification analysis, efficient data preprocessing techniques such as data transformation, data compression, feature extraction and imputation are required. This study investigates three data treatment approaches: randomization; attribute elimination and missing values imputation on bipedal motion data. The effects of data treatment were examined on classification accuracies to retrieve informative attributes. The analysis is performed on bipedal running and walking motions concerning the human and ostrich obtained from public available domain and a real case study. The classification accuracies were tested on seven classifier categories aided by the WEKA tool. The findings show enhancements in classification accuracies for treated dataset in bipedal run and walk with respective enhancements of 3.21% and 2.29% in treated data compared to the original. The findings support the integration of data randomization and selective attribute elimination treatment for better effects in classification analysis.

Simulation of surface roughness and topography in finish turning using digital image subtraction

Most surface roughness prediction models based on mathematical approach assume a geometrically well-defined nose profile, i.e. either circular or elliptical. Such simple models cannot be applied in cases where the tool nose develops different wear patterns or have non-geometric arbitrary profile, such as those caused by manufacturing tolerances in the nose radii. In this study, a digital simulation method using binary image subtraction is used to generate the work piece surface topography, and from which the common roughness parameters of the work piece are determined. Comparison of the digital simulation method with the theoretical models using ideal nose shape shows a maximum difference of 4.7% in the average roughness value.

Feature selection in multimedia: The state-of-the-art review

Abstract Multimedia data mining, particularly feature selection (FS), has been successfully applied in recent classification and recognition works. However, only a few studies in the contemporary literature have reviewed FS (e.g., analyses of data pre-processing prior to classification and clustering). This study aimed to fill this research gap by presenting an extensive survey on the current development of FS in multimedia. A total of 70 related papers published from 2001 to 2017 were collected from multiple databases. Breakdowns and analyses were performed on data types, methods, search strategies, performance measures, and challenges. The development trend of FS presages the increased prominence of heuristic search strategies and hybrid FS in the latest multimedia data mining.

Simulation Approach for Surface Roughness Interval Prediction in Finish Turning

Existing simulation models used in predicting the surface roughness of a workpiece in finish turning are based on an ideal circular cutting tool nose profile. This leads to a single predicted roughness value for a given set of input parameters. In this paper, a simulation approach that considers the random tool nose profile micro-deviations as well as the tool chatter vibration to predict a roughness interval is proposed. The nose profiles used in the simulation were extracted from images of the real cutting tool inserts using sub-pixel edge location. The chatter vibration signal was reconstructed from the measured signals and was superimposed onto the extracted nose profile. The roughness data were computed from 24 simulated workpiece surface profiles and used to determine the 95 % roughness prediction interval. Comparison with the experimental results showed that 100 %, 96 % and 96 % of the Rt, Ra and Rq roughness values obtained experimentally fell within the predicted roughness intervals. (Received in March 2015, accepted in September 2015. This paper was with the authors 1 month for 1 revision.)

Preprocessing compressed 3D kinect skeletal joints in enhancing human motion classification

Human motion classification has been commonly analyzed from 2D and 3D temporal body postures. Previous analyses focused on exergaming, sport science, surveillance and rehabilitation for the betterment of living. A number of recent works had also applied Microsoft Kinect captured motions for its portability and capability convenience to detect RGB images, depth images and 3D skeleton joint coordinates. Nevertheless, Kinect captured data contains inaccuracies which necessitates preprocessing efforts to improve the motion recognition accuracy. To the best of our knowledge, works had rarely detailed the data preprocessing techniques in order to improve the qualities of motion recognition. Few had demonstrated the feasibilities of 2D motion data to represent 3D motions. The importance of data preprocessing role in enhancing the classification accuracy, thus remains a major research concern to date. Therefore, this paper presents the comparisons of human motion data preprocessing analyses on 2D and 3D skeletal joints. Uncertainties observed in raw 3D motion data are identified and filtered followed by the data compression tasks into 2D data. The preprocessing performances at each level are judged using four major classifiers: Bayes, Function, Lazy and Tree aided by the WEKA tool. The approaches are employed on the skeletal joints coordinates of data retrieved from UTKinect-Action Dataset. Our findings demonstrated that preprocessing efforts on the compressed 3D into 2D skeletal joints is comparable to the actual human motion by classification accuracy and at the same time reduces the execution time.

Effect of Tool Nose Profile Tolerance on Surface Roughness in Precision Turning

The effect of the tool nose profile deviations in cutting tool inserts on the surface roughness of the work piece produced based on the actual tool nose profile geometry is studied. The nose profile was detected from the tool nose image captured using the 3-D metrology system. A edge detection approach combining moment invariance operator with Sobel 2-D filter operator is proposed. A work piece surface profile is then generated by considering tool nose profile deviation, feed rate, nose radius and wedge angle to study the effect of the work piece geometry deviation on the roughness values. Based on the experimental results, the maximum differences from ideal and experimental results are 19.8% for R t, 19.9% for R a and 16.1% for R q respectively.