Implementing Accelerometer based measurements

[sparkweaver]

Background

After the problems with using the geolocation class, I looked for different means to do the distant measurement. There are two ways to do it after a quick search

• Accelerometer: Seems fun to fiddle with.
• Using third-party API: This is the most streamlined way but less fun.

Conclusion

I will attempt to use the accelerometer to measure the total distance, but this will be done in a roundabout way.

1. Step Counting: Multiple videos explain how to use the accelerometer, including how to determine when the user takes a step. So, I will include this in the initial design.
2. Distance Estimation: After the step counter works well enough, I will see if I can use the device's measured acceleration with some basic equations to estimate the distance.

[sparkweaver]

Implementation

I have created two classes:

• AccelerometerService: This service takes inspiration from the previous GeolocationService in that it uses an isListening value to prevent repeated subscriptions. Otherwise, it has a Start and Stop listening method where it subscribes and unsubscribes to the Accelerometer.ReadingChanged event. It then forwards the event through its own event, which is planned to be used in different classes later on.
• StepHandler: This utility class update will contain the logic to determine if the accelerometer data indicates a step or not by subscribing to the AccelerometerService.ReadingChangedEvent. The class uses the THRESHOLD constant of 1.2 (current value).
• *UI Integration: Currently, the estimated number of steps is displayed in the middle of the screen.

Testing

When testing the application during a walk, I could clearly see that the number rose by 2-10 units at each step I took. This shows that the initial integration works but needs adjustments to add one unit.

Next Step

There are multiple ways to handle this. I will most likely try the first things that come to my mind:

• Smoothing: Compute an average of the data to get a single peak and not multiple.
• Breaker: Use a timer to break the adding of units.

Sources

• The source for implementing the accelerometer: https://learn.microsoft.com/en-us/dotnet/maui/platform-integration/device/sensors?view=net-maui-8.0&tabs=android
• One of the multiple videos on how to determine steps: https://www.youtube.com/watch?v=XSZBOCpYK7k

[sparkweaver]

Improvements to Previous Implementation

To enhance the step detection accuracy, the following modifications were implemented:

• Refactored: The code in AcceleratometerService was refactored, replacing StepHandler with StepDetector.
• State Memory: StepDetector incorporates a model storing essential parameters for dynamic step detection. This model persists across sessions, providing a personalized user experience.
• Sliding Window: Utilizes a sliding window algorithm to maintain recent data, enabling the computation of a dynamic threshold based on a consistently sized data window and queue data structure.
• Online Update: Maintains two variables, sum and sumSq, updated continuously as new data enters the window. These variables support the real-time computation of the dynamic threshold.

Testing

The application was then tested by myself and a family member, representing fast and slow walking speeds, respectively, to ensure versatility:

• Myself: Demonstrated an improvement over previous tests with variations between 0-3 steps detected per actual step taken.
• Family Member: Yielded similar results, confirming the application's adaptability.
• Both: Initially, the application fails to measure steps and erroneously detects movement when standing still.

Conclusions

• The detected number of steps is still higher than the actual steps taken.
• Inaccuracies are prevalent at the start of a session.
• When stationary, the application mistakenly adjusts to zero movement and records steps.

Next Steps

• Implement a timing mechanism to control step counts within specified intervals.
• Introduce a default threshold to mitigate edge case inaccuracies.
• Establish a minimum threshold to prevent false detections when stationary.

Considerations

Enhancing filtering and control mechanisms impacts the device's performance and energy consumption. future developments may need to address potential compromises.

Sources

• The concept of sliding window: https://www.geeksforgeeks.org/window-sliding-technique/
• The online method: https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance

[sparkweaver]

The feature continued to grow and will contain multiple related challenges. To better manage the development and provide a clear overview of the progression. I have decided to close this issue and split the parts into two new issues:

1. 19 will mention the creation of the step counter to its initial state.

2. 20 will encompass multiple aspects that will be used to enhance the step counter.