Other

Low Pass Filter

A low-pass filter removes fast, high-frequency noise from accelerometer data, leaving only the slower, meaningful movements.

Why use it?

Reduces Noise: Removes small, rapid vibrations or device interference.
Highlights Movements: Keeps the important trends, like walking or posture changes.
Improves Analysis: Leads to cleaner data for calculating features and recognizing activities.

In short, it smooths the data, making movement patterns clearer and more reliable.

Scaler

A scaler adjusts data to fit within a specific range or scale, making it easier to compare and process. It ensures all features have similar influence, especially when they vary in units or magnitudes.

Types of Scalers

Standard Scaler: - Transforms data to have a mean of 0 and a standard deviation of 1. - Helps when data has outliers or is normally distributed.

Min-Max Scaler: - Rescales data to fit within a specific range, usually 0 to 1. - Ideal when you want all values normalized without distorting relationships.

Why Use It?

Improves Model Performance: Prevents large values from dominating smaller ones.
Speeds Up Training: Scaled data helps algorithms converge faster.
Standardizes Input: Makes data easier to interpret and compare.

In short, scalers bring all features onto a similar scale, ensuring fair and efficient analysis!

What is the Sliding Window Technique?

The sliding window technique splits continuous data into overlapping or non-overlapping segments (windows) of a fixed size, moving step by step through the data. Each window contains a small portion of the data for analysis.

How Does It Work?

Window Size: Choose the duration of each window (e.g., 5 seconds).
Step Size: Define how far the window moves forward (e.g., 2 seconds).
Overlap: If the step size is smaller than the window size, windows overlap.

Why Use It?

Extract Local Features: Analyzes smaller sections to capture changes over time.
Handles Long Data: Breaks continuous streams into manageable pieces.
Improves Accuracy: Helps detect patterns, like movement or activity, in specific time frames.

Example:

For accelerometer data recorded at 25 Hz: - A 5-second window contains 125 data points. - With a 2-second step, each window overlaps the next by 3 seconds.

In short, the sliding window technique organizes continuous data into chunks, making it easier to analyze trends and extract features!

What is Magnitude in Feature Extraction?

Magnitude measures the overall intensity of movement by combining the x, y, and z axes of accelerometer data into a single value.

Why Use It?

Simplifies Data: Turns 3D motion into one easy-to-understand value.
Focuses on Intensity: Ignores direction, highlighting the strength of movement.
Reveals Patterns: Helps identify activity levels or motion changes.

In short, magnitude captures how strong the movement is in a simple, direction-independent way.

What is PCA?

PCA (Principal Component Analysis) is a technique used to reduce the number of features in a dataset while keeping the most important information. It transforms the data into new features called principal components.

Why Use It?

Reduces Complexity: Simplifies datasets by combining features without losing key patterns.
Highlights Variability: Focuses on the directions where the data varies most.
Speeds Up Analysis: Makes models faster by reducing the number of inputs.

How Does It Work?

PCA identifies patterns in the data and creates new features (principal components) that capture the largest trends. These components are sorted by importance, so you can choose how many to keep.

Example:

Instead of analyzing 100 features, PCA might reduce it to 10 components, keeping most of the information.

In short, PCA simplifies data, making it easier to analyze and process while keeping the important details.