Research Projects

The global elderly population is already facing challenges of physical and mental well-being that only increases in the upcoming years. Yoga, as already known, channelizes both body and mind towards. The practice of yoga energizes body and helps in overcoming various ailments. However, proper caution is extremely necessary while practicing yoga, since wrong postures may lead to severe complications, particularly in elderly population. The physical trauma may complicate their routine for the rest of their lives. Even strictly following certain postures can be threatening for them. Relaxation to norms are usually handled by tutors in offline guidance. These relaxations are offered to only certain postures and few poses may still require strict evaluations. Towards this, AI guidance may help in assisting the elderly population by calibrating guidance system to every individual. The system would monitor progress regularly by routine reassessment of conditions of the individual, which is useful for people who may not have access to an expert. This system can also be used to consult an expert remotely. The research plan comprises modules like, acquisition of posture templates, mapping muscles involved in each postures using domain information, curating postures by health conditions, estimating flexibility to perform curated postures, suggesting relaxations, and storing vital information to be shared with experts. These are clustered into three stages: the development stage, the testing/validation stage, and the deployment stage. The first stage involves acquisition of data using a camera system. For this, yoga-gurus will be invited to capture template postures and subsets of necessary curated postures will be prepared in their consultation. The relaxations of the respective postures are also recorded for encoding into technical pipeline. The recorded data forms ground-truth for our system with appropriate annotations and meta-data. These will be used to train deep-architectures. Two independent networks will be developed: (1) to classify a pose, and (2) to assess correctness of given posture. This step would also consider motion characteristics and stability of fragments. In the second stage, the developed models are individually validated for performance assessment. After achieving the desired level of performance of individual models, all modules are integrated and tested for an end-to-end benchmarking in the third stage. The developed system is deployed in a controlled space to verify its efficiency and safe operation in a practical world. Another set of yoga practitioners and students will be volunteered in this stage of the project. Practicing elderly people will also be included here to understand future scopes and improvement grounds of the system. All these verifications will be performed by following protocols that mandate necessary permissions, in presence of primary health providers.