M. Masudur Rahman

Human motion recognition using an eigenspace

This paper describes a method for representing and/or recognizing human motion using an eigenspace analysis, referred to a tuned eigenspace, that is responsible for storing various human motions in terms of their sequential postures in an eigenspace and for recognizing unfamiliar posture and/or motion from the tuned eigenspace. We have also analyzed human dress texture problem, caused by wearing clothes, in the proposed method. The performed experiments include representation and recognition of (i) 6 actions of a cricket umpire played by 22 persons wearing respective dresses, and (ii) a turning motion given by a particular person wearing 10 typical clothes. The experimental results show a satisfactory performance of the proposed method for representing and recognizing human posture and/or motion overcoming the dress problem.

Recognizing human behavior using universal eigenspace

An efficient application of eigenspace technique to recognize human behaviors is described. The present paper investigates two types of recognition, i.e., recognition of an unknown human posture and a particular behavior by identifying human postures among several behaviors. A number of different posture sets from some selected behaviors create universal eigenspace and different sets of unknown postures are recognized from it. In contrast to the classical method, the paper proposes to employ some image processing of input images for better performance of the eigenspace technique instead of using just original images for human postures recognition. A new approach for producing eigenspace is described and the robustness of the method is effectively proved in the experiment.

A tuned eigenspace technique for articulated motion recognition

In this paper, we introduce a tuned eigenspace technique so as to classify human motion. The method presented here overcomes those problems related to articulated motion and dress texture effects by learning various human motions in terms of their sequential postures in an eigenspace. In order to cope with the variability inherent to articulated motion, we propose a method to tune the set of sequential eigenspaces. Once the learnt tuned eigenspaces are at hand, the recognition task then becomes a nearest-neighbor search over the eigenspaces. We show how our tuned eigenspace method can be used for purposes of real-world and synthetic pose recognition. We also discuss and overcome the problem related to clothing texture that occurs in real-world data, and propose a background subtraction method to employ the method in out-door environment. We provide results on synthetic imagery for a number of human poses and illustrate the utility of the method for the purposes of human motion recognition.

Robust appearance-based human action recognition

An automatic human action representation and recognition technique is proposed in this paper. Appearance-change problem due to human wearing dresses and body shapes is also investigated in this study for automatic human action recognition. A tuned eigenspace technique is proposed for automatic human posture and/or motion recognition that successfully overcome the preceding problems. We employ image pre-processing by Gaussian and Sobel edge filter, called the first stage tuning, for reducing a dress effect, and a mean eigenspace produced by taking a mean of the similar postures, called the second stage tuning, for avoiding the preceding problems. An eigenspace called a tuned eigenspace is obtained from the mentioned processes and it is used for further recognition of unfamiliar postures and actions. The proposed method is compared with a related technique and the robustness of this approach is presented.

Solving a dress problem for a human model recognition

An approach to recognize a human posture automatically is presented in this paper along with the investigation of a dress effect due to change of wearing dresses. The approach uses an appearance-based eigenspace technique, and augments it by employing blurred-edge operations over original images for reducing dress effect in eigenspace. The proposed blurred-edge operations confirm the expected recognition rate overcoming the dress effect, and related experiment is performed employing a human model wearing various dresses.

Appearance-based representation and recognition of human motions

This paper employs an appearance-based eigenspace technique for representing and recognizing various human motions in terms of their postures. Various human motions are captured by video camera and they are sampled into defined image frames. These sequential images create a multi-dimensional eigenspace in which the motions are represented based on their postures change and this eigenspace is used for recognizing unknown postures/motions based on a MDL (minimum description length) principle. Since human clothes have an influence on creating an eigenspace, we employ blurred edge images throughput learning and training stages instead of original gray-scale images. The proposed method provides eigenspace updating when a new observation becomes available. Experimental results show satisfactory performance on representing and recognizing various motions and/or postures in the proposed approach.

Automatic acquisition of object using eigen appearance

Automatic acquisition of an object using model appearance from an environment is proposed in this paper. Robots directly interact with a defined environment in order to extract object shape from its scene. The robot extracts the targeted objects appearance, creating an eigenspace, and stores it into the memory server (or intelligent data carrier). Eigenspace is constructed every time a new object appears, and various appearances are accumulated gradually. A closed sequence of appearances is generated from the accumulated shapes, which is used for object recognition. Experimental results of object accumulation and recognition show the effectiveness of the proposed method.

HUMAN POSTURE RECOGNITION: EIGENSPACE TUNING BY A MEAN EIGENSPACE

This paper investigates an appearance-change issue due to various human body shapes in an eigenspace analysis, which is responsible for generating person-based eigenspaces employing a conventional eigenspace method. We call this a figure effect in this study for this phenomenon. As a consequence, an appearance-based eigenspace method cannot be effective for recognizing human postures with its present available formulation. We propose to employ a generalized eigenspace for avoiding this problem, which is developed by calculating a mean of some selected eigenspaces. We also investigate a dress effect due to human wearing clothes in this paper. The study proposes image pre-processing by Laplacian of Gaussian (LoG) filter for reducing the dress problem. Since the proposed method tunes a conventional eigenspace as an appropriate method for human posture recognition, the proposed scheme is known as an eigenspace tuning. An eigenspace called a tuned eigenspace is obtained from this tuning scheme and it is used for further recognition of unfamiliar postures. We have tested the proposed approach employing a number of human models wearing various clothes along with their different body shapes, and the significance of the method to human posture recognition has been demonstrated.

OVERCOMING DRESS EFFECT IN EIGENSPACE

This paper addresses a problem, due to human wearing clothes, in an appearance-based eigenspace technique called a dress effect in case of recognizing human postures. The dress effect in the eigenspace arises when a man/woman wears different texture-dress between training and testing stages. This study investigates that change of dresses creates the cloths oriented eigenspace and it affects the robustness of an eigenspace technique. The dress effect is investigated in this paper employing various dress environments and image pre-processing by Sobel-edge filter is proposed in this paper for overcoming this problem. The application of Sobel-edge images over training and testing images creates an almost similar eigenspace of different models/persons, even changing the clothes. Promising obtained results are shown in terms of a numerical analysis, with a graphical representation of various eigenspaces and recognition rates in the performed experiments.

Body shape variability problem in Eigenspace

This paper claims that human various body shapes create individual eigenspaces; as a result, the classical appearance-based model cannot be effective for recognizing human postures. We introduce figure effect in the eigenspaces due to different human body shapes in this particular study. The study proposes an organized eigenspace tuning method for overcoming the preceding problem. Since the proposed method tunes the classical eigenspaces for human posture recognition, we define this phenomenon eigenspace tuning. Generation of a tuned eigenspace (TES) is an organized method where some of similar eigenspaces are selected according to MDD (minimum description deviation) method and a mean if them is taken. In fact, the TES is an optimized visual appearance of various human models that minimizes the fluctuation of MDL (mean description length) between training and testing feature spaces. We have tested the proposed approach on a number of human models considering their various body shapes, and significance of the method to the recognition rates has been demonstrated.