Smart Home Dataset with OpenSHS - Activity Recognition and IoT Sensors

Dataset Overview

The OpenSHS (Open Smart Home Simulator) dataset provides a rich collection of sensor activations from a simulated smart home environment. With 29 strategically placed sensors monitoring various household objects and areas, this dataset captures realistic patterns of daily living activities, making it invaluable for activity recognition and smart home automation research.

Key Features

• 29 IoT sensors distributed across home environment
• Multiple sensor types: motion, contact, pressure, and appliance sensors
• Labeled activity data for supervised learning
• Six activity categories: eat, sleep, work, personal, other, and anomaly
• Timestamp data for temporal pattern analysis
• Realistic daily living scenarios and behavioral patterns
• Anomaly labels for unusual activity detection
• Suitable for both classification and sequence modeling

Data Structure

The dataset is organized with the following structure:

• Sensor Placements: Carpets, doors, lights, bed, couch, fridge, oven, TV, wardrobes, and other household items
• Sensor Headers: 29 binary columns representing sensor activation states
• Timestamp Column: Time of sensor activation with aggregated readings
• Activity Column: Labeled activities including:eat (dining activities), sleep (bedroom/rest activities), work (office/study activities), personal (hygiene/grooming), other (miscellaneous activities), anomaly (unusual patterns)
• Activation Patterns: Sequential sensor firing representing movement and interaction

Data Collection Method

The dataset was generated using the OpenSHS simulator, a hybrid open-source 3D smart home simulator specifically designed for IoT and machine learning research. The simulator combines interactive and model-based approaches to produce realistic sensor activation patterns that mirror real-world smart home environments. Data aggregation was performed during the simulation process to ensure temporal consistency and logical activity sequences.

Research Applications

• Activity recognition and classification systems
• Activities of Daily Living (ADL) monitoring for elderly care
• Anomaly detection in smart home environments
• Behavioral pattern analysis and prediction
• Context-aware smart home automation
• Health monitoring through activity tracking
• Energy-efficient automation based on occupant behavior
• Privacy-preserving activity inference

Machine Learning Use Cases

• Multi-class classification for activity type identification
• Sequential modeling (HMM, RNN, LSTM) for activity prediction
• Anomaly detection using unsupervised and semi-supervised learning
• Time series analysis for behavioral pattern discovery
• Feature selection to identify most informative sensors
• Transfer learning across different home configurations
• Real-time activity recognition with streaming data
• Ensemble methods combining multiple sensor modalities
• Deep learning for complex activity recognition