WEARABLE SENSORS, STEP BY STEP
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
WEARABLE SENSORS, STEP BY STEP
Please read this whole document carefully before starting the coursework, as it contains essential information.
COURSEWORK 2021/2022
Your company is in the process of developing a novel product to monitor the physical activity of their customers. The product is being designed as a wearable device (a smartwatch) and will include a variety of sensors to measure physical activity. As a part of their early-stage research, the company is investigating which sensors work best to monitor their target measurements and which ones are compatible with their product requirements. To this purpose, the company carried out an exploratory data collection campaign and recorded data from a variety of candidate sensors that they could choose from. The company is asking you, their Product Development expert, to investigate which combination of sensors is the most suitable for their product. They want you to find out the best combination of sensors that allows for the longest battery life, while retaining the highest possible accuracy.
You will be provided with the sensors data and will need to write a MATLAB code to estimate the parameters requested by your company. Using the results from the code, you will need to write a short executive report to summarise your main findings.
The deadline for the submission, including MATLAB code and report, is the 9th of December at 16.00.
THE DATA
You will be provided with two sets of data, recorded from the same subject, on two different days (Day 1 and Day 2). The data was collected while the subject was walking on the streets of Bristol, therefore you can expect segments of walking, alternated by segments of no motion (e.g., waiting at a traffic light) and a variable peace. The data was recorded with a smartphone and will include measurements from a variety of sensors (e.g.,
accelerometer, gyroscope, magnetometer…). For a detailed description of the sensors, please refer to the
documentation ofAndroSensor.
Using this data, the company is asking you to estimate:
• The total number of steps
• The total distance walked
• The calorie expenditure
In addition to the sensor data, and only for Day 1, you will also be provided with the ground truth for the three target measurements. You can use this additional data to calibrate your algorithm and set up parameters like stride length and calorie expenditure per step. For Day 2, you will only be provided with the sensor data. You can tune your algorithm to work on Day 1, but the exact same algorithm must be used to process Day 2 as well.
THE SUBMISSION
The final submission must include your source code and the final report. The source code needs to include your algorithm to measure the three target measurements (number of steps, distance walked and calorie expenditure). The algorithm must be executed in the exact same way of both Day 1 and Day 2 (i.e., no tweaking of parameters for different days). A template for the source code will be provided on Blackboard.
In your report, you’ll need to include the results for the target measurements both for Day 1 and Day 2, as produced by your MATLAB code. You’ll also need to include your recommendation for the company, with a brief discussion of your choices and their impact on accuracy, robustness and battery life . You may use as many sensors as you want, but keep in mind that you’re developing a commercial product, and different sensors can have a strong impact on the final product.
The report must be one page only. Additional material can be provided in form of figures and tables, but it must fit in the same page. References are not required, but if provided, they can be added to a second page.
FURTHER READING
The steps to produce a basic algorithm will be discussed during the example classes. For further reading on more advanced techniques, you can refer to the following paper:
Salvi et al., An Optimised Algorithm for Accurate Steps Counting From Smart-Phone Accelerometry (link)
2021-12-05