A Dataset of Inertial Measurement Units for Handwritten Maths Numbers and Symbols

The dataset consists of Inertial Measurement Unit (IMU) data corresponding to the 10 numeric digits (0–9) and 4 mathematical symbols: plus (+), minus (−), multiplication (×), and division (÷). The data was collected using an IMU 6050 sensor, which was attached to a marker held by participants during the handwriting process. The IMU sensor captures accelerations along three axes (X, Y, Z) and rotational velocities along the same three axes, providing detailed motion profiles for each written symbol.

Data Collection Process:

Twenty students participated in the data collection process for this study. Each student was tasked with writing all 14 characters (10 digits + 4 symbols) twice — once with the IMU sensor attached to the upper part of a marker and once with the sensor attached to the lower part. This dual positioning helped investigate how sensor placement affects motion data distinctiveness and recognition performance. Each student thus contributed 28 samples (14 characters × 2 sensor positions), creating a balanced and diverse dataset.

The data collection process was conducted over a four-month period to ensure ample data diversity and volume. In each session, students wrote each symbol 250 times on a whiteboard. To precisely record the start and end of each character-writing event, participants used a button they pressed before starting and released upon completion. The IMU sensor continuously recorded data for all 250 repetitions of each character written by each student. This approach resulted in a rich, large-scale dataset well-suited for deep learning and machine learning applications.

Labeling:

The data is labeled numerically from 0 to 13. Labels 0–9 correspond to digits '0' through '9', label 10 represents the plus sign '+', label 11 represents the minus sign '−', label 12 represents the multiplication sign '×', and label 13 represents the division sign '÷'. This straightforward labeling allows for efficient classification and identification during analysis and modeling.

Data Interpretation and Usage:

Character Recognition: Train and evaluate machine learning models to recognize numeric digits and basic arithmetic symbols based on IMU data.

Sensor Analysis: Investigate how different sensor placements affect recognition accuracy and develop methods for sensor position normalization or compensation.

Handwriting Dynamics: Analyze the writing speed, stroke patterns, angular velocity, and other motion features associated with handwritten digits and symbols.

This dataset offers a valuable resource for researchers and developers working on handwriting recognition using inertial sensor data, particularly in the context of numeric and symbolic input recognition.