Method for assessing the similarity and distance between curves/trials and its applications in gait analysis

In clinical assessment or sports exercise, it is common that a subject is required to repeat a specific performance so that a stable movement pattern is obtained and analysed. In practice, however, the trials done by a subject vary more or less, depending on the psychological or physical conditions. Some of the trials can be used as representatives of the stable movement pattern, and some not. Therefore, there is a demand for a new method to identify which trials/curves are similar. The traditional methods used to assess curve similarity are not perfectly suitable for the case where there are only a few of trials available. This study proposes a similarity-distance coefficient to assess the similarity of curves/trials. A group of designed curves are used to validate the coefficient. The results show that given joint kinematic data during gait as examples, the proposed coefficient can be used to quantitatively evaluate the similarity of trials, and thus find which trials would be representative (s) for the gait investigated. The proposed method could be applied in various situations where repeat movements have to be measured and analysed.