The purpose of this paper is to present the design and implementation of a novel rule-based algorithm for the classification of sitting postures in the sagittal plane. The research focused on individuals with severe musculoskeletal problems and, thus, specific requirements for posture and pressure management. Clients' body shapes were captured using the Cardiff Body Match system developed by the Rehabilitation Engineering Unit, Cardiff and Vale University Health Board. The algorithm consists of four main steps: the first step is the symmetry line detection, the second step involves the mathematical analysis of the curvature of the backrest profile, the third step is the sitting posture classification and the fourth step is the extraction of the geometric parameters from the curve. The results show the classification system was successful in identifying four types of curves characterizing sitting postures using local derivatives as curve descriptors with an overall accuracy of 93.9%.