It is January 2012, and Jacques Kallis faces Dennis Lillee at Sabina Park cricket ground. He hooks a bouncer. The ball he faced just before was against Courtney Walsh, which he played with a forward defensive stroke. Yet Jacques is not in Jamaica, but in the Virtual Reality Cricket Research Laboratory at the University of the Witwatersrand (South Africa).
To help players like Jacques improve their technique, we are developing a batting analysis and training program and a virtual-cricket-ground environment. The system improves the execution of batting strokes and provides an interface to the virtual environment. It captures the motion of a batter as he plays a stroke, then compares it to known strokes and provides feedback outlining how well the play was executed. Our trainer and analyzer allow us to compare a cricketer's action to a standard reference. The reference comes from several sources, including a coach or trainer's descriptions of the motion, coaching literature,1–3 and the movement of more experienced batters.
Cricket's objective is to hit the ball with a bat around the ground to score runs. The speed of the swing varies with each batter and helps determine the stroke's effectiveness. To play well, a cricketer must gauge the position of the bat at a particular instant in time. Players should consider other factors as well, such as the pitch and speed of the ball, its bounce height, and its lateral movement in relation to the pitch. An effective training system must weigh all of these elements. Our system analyzes the player's current motion and technique and suggests corrections. It also monitors the cricketer's implementation of these corrections.
The ability to reliably reproduce the same motion is very important in cricket and an expert batter is more effective than a novice in identifying different types of delivery. We designed the system to give a novice the benefit of an experienced player's perception.
The player is taken through the sequence of motion throughout the dynamics of the ball,3 using both video analysis and practical demonstration. The coach then sets up the conditions and the system evaluates the stroke's execution.
Other constraints such as the pitch, the weather, the form of both the batter and bowler, the fielders' positions, and the score must also be taken into account by the player and training system to yield optimum performance.
The trainer program must give the cricketer a clear illustration of vague body-posture terms such as foot forward, head up, or head level. To accomplish this, fuzzy membership functions relate angular variables, as well as length or positional variables, to these coaching directives. Each posture, such as front foot forward, head level, or head down, has a corresponding membership function, which can be generated from Figures 1–3.
Figure 1. (top) Vertical and (bottom) horizontal head positions.
Figure 2. (top) Back foot and (bottom) front foot positions.
Figure 3. (top) Angular and (bottom) linear bat and feet positions.
Our system separates the cricket stroke into stages of preparation: back swing, forward swing, impact, and follow through.4 We divide this motion into three time intervals: the waiting, receiving, and playing/follow-through states.The waiting state describes the batter taking up position at the crease to face the bowler. The receiving state starts when the bowler releases the ball and the batter moves to find the best position to play the selected shot. During this state, the batter seeks postural cues from the bowler to perceptually anticipate the lateral displacement, pitch, and speed of the ball.5 The playing and follow-through state includes the ball making contact with the bat, the orientation of the bat to play the stroke, and the player's motion after the ball leaves his bat.
Figure 4. Position of sensors on batter (batter at the crease).
The beginning and end of the motion is important for the segmentation. We reference the start of the motion to the time the ball leaves the bowler's hand. Therefore, the waiting state starts at a time (t1) before this point and ends at this point. The receiving state (t2) begins from this point and extends until the ball makes contact with the bat. The playing and follow-through state (t3) starts when the ball makes contact with the bat and ends at time T after the ball leaves the bat and the entire motion for the stroke has ended. Figures 1–3 show the range of motion for the head, feet, and bat during these four temporal states.
We used a Polhemus Fastrak system that provides the, x, y, and z coordinates of the 3D position in space, and orientation information in yaw, pitch, and roll (see Figure 4).
In summary, our cricket-batting-technique analyzer/trainer uses fuzzy-set theory to categorize batting strokes. Our ongoing work is classifying batting strokes and integrating the system with a virtual cricketing environment.
School of Computer Science
University of the Witwatersrand
Johannesburg, South Africa
Mervyn Curtis is the head of school. Prior to this, he was the head of the Department of Mathematics and Computer Science at the University of the West Indies, the head of the School of Engineering in Jamaica, and a senior lecturer in the Department of Electrical and Electronic Engineering at the University of Nottingham (UK).