User-Independent Intent Recognition on a Powered Transfermoral Prosthesis
Abstract
Powered prostheses are a promising new technology that may help lower limb amputees to function at higher levels in their daily lives. There are over 650,000 individuals in the United States who have suffered from major lower limb amputation (loss of leg above the ankle). These individuals suffer from significantly impaired mobility including expending up to 60% more energy than non-amputee individuals. Less than 25% of transfemoral amputees older than 50 achieve community mobility on passive prostheses. Research and industry teams have begun building powered prostheses that include motors to actively assist amputees to walk and perform various tasks encountered in everyday situations such as stepping up a stair, standing up, and traversing difficult and uneven terrain such as slopes and ramps. These devices are controlled by an onboard computer that determines the timing and magnitude of assistance to help an amputee. An important objective is for the computer on the prosthesis to understand what the amputee wants to do. By accurately decoding the amputee s intentions, the computer can appropriately coordinate the assistance of the powered prosthesis to the amputee s needs. A powerful technique to understand the amputee s intentions is to use pattern recognition, which is a technology that is commonly used in speech recognition, image analysis, and medical diagnostics. Pattern recognition is capable of automatically determining the amputee s intent and can allow amputees to easily and intuitively use their powered prostheses in their everyday lives. However, if the pattern recognition software incorrectly estimates the user intent, then the powered prosthesis may not be as helpful or may even get in the way of an amputee s intended movements. Additionally, pattern recognition requires training data that must be collected from the amputee before using it. We have developed new pattern recognition systems that are more accurate and do not necessarily require training data directly from the amputee. The proposed research will develop and test these pattern recognition systems with amputees using a state-of-the-art powered prosthesis. The research will determine the benefit of pattern recognition intent recognition systems by measuring key clinical parameters such as how quickly amputees are able to move with the powered prosthesis and their energetic cost of doing so. The end result of this research will be intent recognition systems capable of implementation on computers embedded on powered prostheses. Powered prostheses such as the iWalk and Power Knee are already available, and more devices will be commercialized in the next few years. This will be useful to lower limb amputees who use powered prostheses in the future as intent recognition systems can help amputees achieve a greater level of independence and mobility. This research will be of benefit to U.S. Veterans who have lost one or both legs during service, which is 83% of the major limb amputations that military surgeons have performed in the previous decade. The ability to have community ambulation and independence for an amputee helps not only the patient but also helps to reduce the burden of care on immediate family and caregivers. This research seeks to innovate new ways to ensure that new powered prostheses can be highly beneficial to amputees in the real world.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 31, 2017
- Source ID
- W81XWH1710031
Entities
People
- Aaron Young
Organizations
- Georgia Tech Research Corporation
- United States Army