The Parkinson’s disease (PD) is a frequent chronic progressive syndrome in the elderly population. Current available PD treatments either stimulate brain dopamine receptors or increase dopamine synthesis. In the long-term, especially from the fifth year of disease on, onset of motor complications is often observed. Such motor complications arise as a consequence of a reduction in the duration of the effect of the medication. This motor complications affects spatiotemporal properties during the gait of patients [8]. Length and speed of the steps change due to the effects known as dyskinesia and or akinesia, thus, and on-line measurement of these properties during gait may help to predict these situations and therefore warn the patient (e.g., minimizing risk of falling) or a remote health care system. Recently several approaches using inertial sensors have been developed with the aim of measure these types of gait properties [1, 2, 4, 5, 6]. The gait characteristics may be obtained using gyroscopes tied at legs using a double inverted pendulum model [1, 5], nevertheless wearing these devices on the legs during daily life activity seems a drawback, leaving the application scope of this method to clinical environments. In the case of accelerometers, they are usually positioned at the dorsal side of the trunk near the L3 region of the subject, since it is the CoM location. In this position, 3D CoM acceleration, velocity and displacement can be estimated [2, 4]. However, to the best of our knowledge, there is no a user-friendly wearable device/location, that patients may use outside the hospital. Here we propose a method, based on SVM-regression, to extract spatiotemporal properties from accelerations obtained from a single accelerometer positioned at the lateral side of the waist, with the advantage of being a wearable system the patients may use during their daily life, without danger of hurt or damaging the device.