In this work we propose a model (in)validation approach to gait recognition, using a system that tries to discriminate specific activities of people. The recognition process departs from an abstraction obtained from video image sequences for different activities performed by different people, by first using a suitable representation for each frame and for each frame sequence. For each frame two commonly used models for describing silhouettes are employed: Fourier Descriptors and vectors of widths. Then each sequence is modeled as a linear time invariant (LTI) system that captures the dynamics of the evolution of the frame description vectors in time. Finally a standard classification tool, SVM, is used to recognize activities using similarity measures obtained through model (in)validation. The main contribution of this work is the provision of an activity recognition model and the performance evaluation of this model using two different feature spaces.