Analysis of Human Gait Cycle Using Machine Learning

The main objective of this report is to present the idea of measuring the human gait cycle through a deep learning approach using wearable sensors. As the world is changing every day to advance technologies, it is an utmost requirement of the doctors or researchers who are working on human body mech

Project Title

Analysis of Human Gait Cycle Using Machine Learning

Project Area of Specialization

Mechatronics Engineering

Project Summary

The main objective of this report is to present the idea of measuring the human gait cycle through a deep learning approach using wearable sensors. As the world is changing every day to advance technologies, it is an utmost requirement of the doctors or researchers who are working on human body mechanisms and their phases to have the latest techniques that make their work easy and smart. In this context, the deep learning approach provides them a better way of understanding the human gait cycle due to the extraction of only useful data sets for their work. The key challenge of analyzing the human gait cycle is the continuous data that is provided by wearable sensors like IMU, and different FSRs. The most important challenge in this field is to collect time-series data over a specific period. As the human gait cycle of each person is different in sense of their motion and walking cycles, many parameters remain undetected if the gait cycle data is provided over a long period. To tackle all these problems, we are proposing a deep learning approach using an algorithm of artificial neural networks. To tackle the time-lag and vanishing gradient problem, Long-Short Term Memory (LSTM) recurrent neural network can be used as it works on different layers of input and output for the extraction of data set series. Here python library Tensor Flow will develop an intelligent approach that can be used according to the user’s need. We will develop a model using wearable sensors and the optimization of their data sets using LSTM for the best outcomes. It involves the analysis of the human body gait using a different algorithms for the processing of data over a specific time series. Our approach will provide a solution to the problem of vanishing gradient and hence provide a useful method for the doctors and researchers working in this field. From this technique, we can measure the variation of the gait cycle from person to person using different algorithms and fusion of neural networks. In this project we will be using published as well as custom data sets. For custom data sets we are using a combinations of IMUs and strain guages  for the detection of gait phases.

Project Objectives

In this project, we are proposing an idea in which:

• We are acquiring data through sensors • We are testing the algorithms on custom as well as on published data sets.

• We are using LSTM recurrent neural network to analyze gait pathologies over a long period of continuous time series.

• We are going to achieve mean normalized curve of human gait cycle

Project Implementation Method

Human gait is a cyclic process and is defined by different key events like heel strike, toe off which are subdivided into swing and stance phases. To measure all these phases we use a set of strain gauges on the foot. IMU is used for determining joint angles and orientation and also for measuring flexion and extension of the lower limb. IMU will provide us the recognition of walking sitting and standing positions. Strain guages will provide us the data for heel strike, toe off, midswing and other sub-phases of human gait. After acquiring all the training data sets we will give this information as input to LSTM which flows through different layers like the input layer, hidden layer, and output layer of LSTM. LSTM will process this data over a continuous time series and solves the problem of vanishing gradient and correlation of different time-lag problems. For the proposed architecture input data is fed directly into the input layer, then the full output of the first layer is fed into the next layer as input, and so on... After the output is obtained from the last output layer of LSTM, these output values can be used to plot the mean normalized curve of the human gait cycle. In the end, we will predict/analyze the accuracy of the proposed system by comparing it with other published data sets.

Benefits of the Project

The benefits of our project are:

Measuring the human gait using strain guage and IMUs will provide better results.

AIgorithm of LSTM will provide better a way to handle time series data.

A standard of mean normalized curve for gait cycle will be provided.

Technical Details of Final Deliverable

By the end of this project, the end-products that are deliverable to costumers are:

• Gait analysis system

• Algorithm of LSTM

• Data sets

• Research Article

• Mean normalized curves for gait cycle

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Medical

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Good Health and Well-Being for People, Quality Education

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com