Lower-Limb Rehabilitation at Home: A Survey on Exercise Assessment and Initial Study on Exercise State Identification Toward Biofeedback

Introduction

Disability greatly affect the quality of life of people all over the world. The world report on disability stated that the number of people having a disability will increase to over one billion people (World Health Organization, n.d.-b). The rising number is due to the fast ageing population, chronic health conditions and injuries. Rehabilitation is the compulsory procedure for recovery and prevention from further disabilities so as to reduce the risk of fall and hospital admission for the elderly (World Health Organization, n.d.-a) as well as to prevent and treat functional disability (Singh, 2002). Lower-limb injuries that need rehabilitation include osteoarthritis (Musumeci et al., 2014), knee injuries (orthoinfo, n.d.), Dynamic Knee Valgus (DKV) among althletes (Gornitzky et al., 2016) and post-reconstruction surgery (Paterno, 2015) etc. While rehabilitation is proven crucial, injury prevention during recovery is often overlooked. Thus, the patients are encouraged to have more regular supervision.

While clinical physiotherapy diagnoses and customized treatments often rely on the visual assessment of experienced physiotherapists, technology tools such as motion tracking system and wearable sensors have facilitated quantitative progress reporting and biofeedback for patients practicing at home. It is compulsory that the patient repeats the exercises as instructed during the clinic appointment. If patients fail to execute the correct motion or improper placement of the lower limbs while exercising, it would slow down the recovery progress or place unnecessary pressure on the joints concerned (Draper & Ballard, 1991). One consequential example of mis implementation is Dynamic Knee Valgus (DKV) (Padua, Bell, & Clark, 2012) which is a result of putting pressure on the knee joint by twisting it inward. It could escalate to Anterior Cruciate Ligament (ACL) rupture. Enabling assistive technology maximizes the accessibility and affordability of home rehabilitation, especially at the community level. In tandem with solving the issue of inadequate number of physiotherapists and caregivers for the ageing population, digital supervision of physiotherapy exercises are feasible and scalable solutions.

Wearable sensors, Internet of Things, and available devices serve to deliver clinical and quality health care to people at home or in a community. Realizing the importance of eHealth and the availability of motion tracking system, home-based rehabilitation has become a leading research topic in recent years. The collective information of MEMS sensors not only provides progress summary to the physiotherapist but also has the potential to assess and take earlier measures to prevent injuries while the exercise is being carried out in real time.

In this paper, we propose a machine learning model for human motion assessment for state identification to facilitate rehabilitation exercise assessment. Upon proposing the training model, the survey on state-of-art methods of exercise assessment and state recognition are summarised. The paper will provide the insight and impact of input features, the signal data and sensor-fused angular displacement, align with the performance of the pattern recognition model.