IntroductionCerebral palsy (CP) refers to various motor impairments caused by damage to the central nervous system during fetal development (1). This disorder affects approximately 0.3% of births (1) and often manifests itself during the early childhood as a difficulty to use one side of the body (hemiparesis). Motor deficits encompass difficulty in planning and executing movement.Physical therapy is proposed for children with CP to help them to grow and physically develop as well as possible. The traditional approach of physical rehabilitation focuses on muscular strength and proposes repetitively simple and uncontextualized movement. However, such therapies are of little interest to children and offer limited functional value in daily living, affecting their motivation to continue the therapeutic activities (2, 3). Constraint-induced movement therapy (CIMT) has been developed to improve movement patterns and to maintain the range of the affected arm and leg joints. This approach is often used to improve upper limb function (4). In this therapy, impaired people are encouraged to use their affected hand by restricting the unaffected hand and asking for intensive movement with the impaired upper limb. However, having the unimpaired arm blocked for long periods of time can generate frustration in the child and might not be applicable in a long-term rehabilitation program. Thus, more childfriendly approaches are needed during the neurodevelopment of children with CP.Upper-body interactive rehabilitation system for children with cerebral palsyThe field of virtual reality (VR) has grown dramatically as an emerging tool showing a great potential for use in physical medicine and rehabilitation (5, 6). VR systems have a capability to achieve rehabilitative goals through the use of realtime feedback as well as adaptive strategy and/or difficulty (7-9), and several studies have shown hopeful results; however, few researchers focus on cerebral palsy (CP) (for review see 10). In the same vein, we have begun a technical and clinical project aiming to create an efficient VR-based game to improve the upper limb function of children with CP by encouraging them to use their affected hand as well as to improve their movement and motor control of the limb (11).Virtual Rehabilitation System: Our system for rehabilitation proposes a catching task (see Figure 1). Standing upright in front of a screen monitor, the user can control the upper limbs of a displayed avatar by moving their own upper limbs. A MicrosoftKinectTM sensor was placed at the bottom of the screen, to capture positional data of the user's left/right hands, wrists, elbows, and shoulders in 3D space. Our system maps the data to the movements of the avatar's limbs. The movements of the avatar's hands are represented on a circle displayed at the center of the screen. Positional data is converted to 2D positional data and rendered in real time to provide visual feedback to reduce motor errors. The sample rate of the Kinect sensor was simulated in the catching task. The sample rate data were collected while the catching task is played for 2 minutes. As a result of the simulation, the average sample rate was 20.67 Hz (SD = 2.51). According to the simulated average sample rate, the joint positions were recorded about 20 times per second.Catching Task: In the proposed catching task, virtual objects appear randomly at the border of screen, one by one, and move toward the center. The user controls the virtual limbs of the avatar in an attempt to touch a virtual object moving around within the virtual space. The application emulates multiple properties of the virtual object, including direction and velocity of movement, size or shape. If the user catches one object before it arrives at the center, he or she wins one point. The system supports two interaction techniques: (a) single hand (only the leftor right hand is used) and, (b) both hands (left/right hands are used simultaneously or separately). …