BACKGROUND: Most of the patients who survive stroke, spinal cord or others nervous system injuries, must face different challenges for a complete recovery of physical functional impairment. An accurate and recurrent assessment of the patient rehabilitation progress is very important. So far, wearable sensors (e.g. accelerometers, gyroscopes) and depth cameras have been used in medical rehabilitation for the automation of traditional motor assessments. Combined with machine learning techniques, these sensors are leading to novel metric systems for upper limb mobility assessment. OBJECTIVE: Review current research for objective and quantitative assessments of the upper limb movement, analyzing sensors used, health issues examined, and data processes applied such as: selected features, feature engineering approach, learning models and data processing techniques. METHOD: A systematic review conducted according to the PRISMA guidelines. EBSCOHOST discovery service was queried for relevant articles published from January 2014 to December 2018 with English language and scholarly peer reviewed journals limits. RESULTS: Of the 568 articles identified, 75 were assessed for eligibility and 43 were finally included and weighed for an in-depth analysis according to their ponderation. The reviewed studies show a wide use of sensors to capture raw data for subsequent motion analysis. CONCLUSION: As the volume of the data captured via these sensors increase, it makes sense to extract useful information about them such as prediction of performance scores, detection of movement impairments and measured progression of recovery.