People use gestures to give verbal communication more expression and also to replace the speech. One of the most concise and expressive gestures is the pointing gesture. Pointing gesture can be observed in the early childhood. In these early years, it is used to point at objects or people. Later, people use pointing gestures even for more complex things such as to visualize directions. Increasing pointing gestures are also used for interacting with computers. For example, gestures can be used to remotely interact with a display, without using an input tool. In this work we investigated how people point to objects and how the recognition accuracy can be improved by using a gesture recognition system. We performed a user study, where participants had to point on projected pointing targets. These gestures were recorded as reference data with the help of motion capture system. We used a variety of starting positions in which the study was carried out. For this the participants were placed with a distance of 2 to 3 meters to the pointing targets. At these two positions the participants had to sit and stand while pointing to the targets. From the recorded reference data we derived a pointing vector. Each vector describes the direction in which the gesture is directed. We generate these vectors out of different body parts. This is done to show that there are different ways to create these vectors but they behave all the same. In the optimal case, this vector would describe the path of the person pointing to the object, in this case, the projected point. By mathematical analyzes we show that in average over several experiments and over several participants a systematic deviation from this optimal vector can be detected. We specify models, which can compensate the systematic deviation. These models shift the pointing vector in the direction of the average distance between optimum and average study vector. Products of the consumer market can be used to detect pointing gestures. There gestures can be improved with the generated models. The focus here is, for example, products like the Kinect.