click here to download the abstract project of android
ABSTRACT
We provide android project in this project. Human activity recognition (HAR) aims to develop a mobile application that can accurately recognize and classify various human activities using the sensors available in Android smartphones. The application leverages data from accelerometers, gyroscopes, and other sensors to identify activities such as walking, running, sitting, standing, and more. By applying machine learning algorithms to this sensor data, the app provides real-time activity recognition, which can be utilized in fitness tracking, health monitoring, and context-aware applications. The goal is to create a robust and efficient solution that operates effectively on a wide range of Android devices.
Existing System
In the existing system, human activity recognition is often performed using dedicated wearable devices or through manual logging by users, which can be cumbersome and inaccurate. While some smartphone applications attempt to recognize activities, they typically suffer from limited accuracy due to the simplistic use of sensor data and lack of sophisticated algorithms. Additionally, existing solutions may not be optimized for real-time processing or may drain the device’s battery quickly due to inefficient use of resources.
Proposed System
The proposed system introduces an advanced Android application that employs machine learning techniques for accurate and real-time human activity recognition. The system uses data from multiple sensors available in smartphones, such as accelerometers and gyroscopes, to classify a wide range of activities. The app is designed to be resource-efficient, minimizing battery consumption while maintaining high accuracy. The system also includes a user-friendly interface, allowing users to monitor their activities and track their movements over time.
Methodology
- Data Collection: Collect sensor data from the accelerometer, gyroscope, and other relevant sensors on the Android device. This data includes information on the device’s orientation, acceleration, and movement patterns.
- Data Preprocessing: Process the raw sensor data to remove noise and irrelevant information. This includes filtering, normalization, and segmentation of the data into meaningful intervals.
- Feature Extraction: Extract relevant features from the preprocessed data that are indicative of specific human activities. Features may include mean acceleration, variance, signal magnitude, and frequency domain characteristics.
- Model Training: Train machine learning models, such as decision trees, support vector machines, or neural networks, using the extracted features. The models are trained on labeled datasets that include different human activities.
- Real-Time Activity Recognition: Implement the trained model in the Android application to perform real-time activity recognition. The app continuously collects sensor data, preprocesses it, and classifies the current activity based on the trained model.
- User Interface: Develop a user-friendly interface that displays the recognized activities and provides insights or recommendations based on the user’s activity patterns.
- Performance Optimization: Optimize the application for real-time processing and low battery consumption. Techniques include efficient data sampling, model compression, and background processing.
Technologies Used
Firebase: For cloud storage, user authentication, and real-time database if cloud integration is required.
Android SDK: For developing the mobile application and accessing the device’s sensors.
Java/Kotlin: Programming languages used for Android development.
Sensor APIs: Android APIs for accessing accelerometer, gyroscope, and other sensors.
Machine Learning Libraries: TensorFlow Lite, ML Kit, or other machine learning frameworks for training and deploying models on mobile devices.
Python: For data preprocessing, feature extraction, and model training (prior to deploying the model on the Android app).
SQLite: Local database for storing user data and activity logs.
