Cardiovascular disease (CVD) is the leading cause of premature death and disability in Europe and worldwide. Effective Cardiac Rehabilitation (CR) can significantly improve mortality and morbidity rates, leading to longer independent living and a reduced use of health care resources. However, adherence to such an exercise programme is generally low for a variety of reasons such as lack of time and how enjoyable the CR programme is. In this work, we proposed a method for automatic enjoyment estimation during an exercise which could be used by a clinician to identify when a patient is not enjoying the exercise and therefore at risk of early dropout. In order to evaluate the proposed method, a database was captured where participants perform various of CR exercises. Three set of facial features were extracted and were evaluated using seven different classifiers. The proposed method achieved 49% average accuracy in predicting five different enjoyment level on the newly collected database.