According to the World Health Organization (WHO), musculoskeletal chronic conditions are the leading causes of disability worldwide. Chronic conditions, such as arthritis, have no specific lab tests, and a diagnosis is formed based on a constellation of subjective exams. Physicians also listen to joint sounds to evaluate joint health, scientifically known as joint acoustic emissions (JAEs).  As Thomas R. Insel says, "The good news stories in medicine are early detection, early intervention". To facilitate early detection of joint disease, this work focuses on JAE analysis as a digital biomarker for knee health assessment through machine learning and signal processing techniques using wearable devices. To achieve this, JAEs from loaded and unloaded knees were recorded during the treatment of patients with juvenile idiopathic arthritis (JIA). Machine learning models were developed using JAEs to assign health scores. Model predictions supported the clinical records of successful treatment and showed that the loaded and unloaded exercises contained different and possibly clinically relevant information. To improve the pre-processing and reliability of JAEs a novel algorithm was developed to detect and exclude JAEs contaminated with rubbing artifacts. Then, JAEs were explored under two different loading conditions and compared against synchronously recorded knee biomechanical signals to determine their attribution to the biomechanics, verifying that JAEs contain salient information on knee tribology. JAEs result from friction between various knee surfaces during movement cycles, not all of which are clinically relevant. Those knee phases generating more informative JAEs to distinguish between patients with osteoarthritis (OA) and healthy controls were identified in three routine clinical maneuvers from several locations around the knee. Focusing on certain locations and phases significantly improved classification performance and reduced the computational load. This work lays the foundation for improved usability of JAEs as a quantitative diagnostic biomarker in patients with JIA or OA, and it establishes a strong quantitative correlation between JAEs and knee tribology. The importance of exercise type and microphone placement was highlighted, and informative phases were determined to enhance the computational efficiency of wearable devices.