Context Aware Media Recommendation System
In this era of technology people are facing problem of choices. Technology has allowed us to generate so much data and information that people feel it hard to choose between different kinds of items. We have much diversity in multimedia including movies, music, reading articles and pictures. People
2025-06-28 16:30:55 - Adil Khan
Context Aware Media Recommendation System
Project Area of Specialization Wearables and ImplantableProject SummaryIn this era of technology people are facing problem of choices. Technology has allowed us to generate so much data and information that people feel it hard to choose between different kinds of items. We have much diversity in multimedia including movies, music, reading articles and pictures. People need recommendations and suggestions to choose the right content which must be according to their taste and habits. Many solutions are out there in the market. Some of them are quite sophisticated but they are deployed by word leading platforms for their services instead of available as standalone applications. Amazon uses recommendation algorithms to recommend right items to the buyer. Netflix uses it to recommend right movies. They are some issues with the available solutions: they are either specific to one kind of multimedia or they are not detailed enough to capture more information regarding the users.
In this project we are focusing on a new solution. We are targeting user’s contextual information to create a scenario. Contextual information contains current time, location, weather, and human activity. The scenario allows to choose specific category of multimedia like entertainment, sports, history etc. Based on the chosen category and user’s defined preferences the user is being recommended multimedia including movies, music, reading articles and pictures from the group of larger datasets.
This project consists of both software development as well as hardware integration. Different wearable sensors are utilized to capture user’s contextual information e.g. motion sensors, GPS. Software part consists of mobile application to be used as GUI and server-side implementation to perform the heavy processing so that user’s mobile device is not burdened.
Project Objectives- To develop an understanding of IoT and Pervasive Computing
- To utilize Machine Learning for data processing
- To interface wearable sensors to be used for context recognition with a smartphone
- To design an algorithm to recognize the user’s context based on smartphone and wearable
- sensors
- To develop and Android application for user context recognition and personalized media
- recommendation
Hardware Modules
Different IoT kits are being integrated and wearable sensors are connected together the smartphone through some communication channel.
Data Acquisition
We are using smartphone sensors as well our hardware module to get data from the sensors. Sensors consist of Accelerometer, Gyroscope, GPS. Smartphones have built-in sensor manager which return raw sensor data. Our hardware module on other hand calculate this data and transmit over the Bluetooth to the smartphone. Bluetooth is a low powered wireless communication device configured to transmit data at a rate of 2.45 GHz using Radio Frequency Waves. Typical range is 10m up to 100m.
Pre-processing
Many smartphones nowadays are more efficient than their previous variants. Effects of environment and internal system architecture is much less on the sensors data. As the said effect is less but not negligible. We’ve to perform some pre-processing on the raw sensors data to remove the noise effect. For this purpose, averaging filters or median filters can be used. Further, the data is coming in a continuous fashion but the letter stages require some specified number of samples per unit time. Therefore, the segmentation process is required to achieve the desired task
Machine Learning
ML is then used to train models which predict and give us the human activity as well as provides the information of scenario. Based on generated information another machine learning model recommends the right content.
Server and Client Integration
Heavier tasks like pre-processing and machine learning is performed on server side where it's implemented in open source python frameworks and client (android app) is integrated through http requests. Soft tasks like user preferences handing, recommendations displaying, and rating fetching is performed in android app on client side and passed to the server.
Benefits of the ProjectThis project is related to data science and machine learning. It provides following technical and non-technical benefits to ourselves as well customers:
- A unique solution to the existing problem.
- More detailed and specific recommendations to the individual users as it accommodate contextual information to capture a scenario.
- No more wastage of time on wrong and distasteful content to be consumed. Users can save time and consume right content according to the taste.
- Lightweight application which don’t burden user’s smartphone much and saves phone’s battery.
- Targeting different kinds of multimedia including movies, music, articles and pitcures instead of single.
The project have three final deliverables of this project:
A hardware module as a wearable sensing device:
This module can be worn on wrist which tracks user's motion and allows to identify human activity. The module also contains other controllers which helps in identifying current location as well as weather states based on GPS data. It has Bluetooth communication available to send the data to mobile application and receive status.
Algorithm for user’s context recognition and media recommendation
This part of the project is being implemented in Python framework on a machine running Linux OS. Communication between client mobile application and this server is being performed using HTTP requests. Machine learning method helps determining and making sense of user’s context. It provides way to create a unique scenario. The algorithm utilizes state-of-the-art Matrix factorization method to provide recommendations. Users’ preferences and other information is stored in databases and file storage systems both online (like Firebase) and local (like Mysql).
Android App:
The mobile app act as a GUI. It allows users to set preferences. It displays the identified and calculated recommendations from the server to the user and allows users to rate them for further fine tuning. Mobile app also acts as a bridge between server side and wearable module where it receives data from the module through Bluetooth communication and pass to the server using HTTP requests. It’s a lightweight app because of client-server architecture which is good for both maintain smartphones performance and updating the app itself.
Final Deliverable of the Project HW/SW integrated systemType of Industry IT Technologies Artificial Intelligence(AI), Internet of Things (IoT), Wearables and Implantables, Others, Big DataSustainable Development Goals Good Health and Well-Being for People, Quality Education, Industry, Innovation and Infrastructure, Partnerships to achieve the GoalRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 42120 | |||
| Thunderboard Sense IoT Kit | Equipment | 1 | 10000 | 10000 |
| Arduino Pro Mini | Equipment | 1 | 500 | 500 |
| Rechargeable Batteries | Equipment | 3 | 400 | 1200 |
| Arduino Uno | Equipment | 1 | 700 | 700 |
| Oled | Equipment | 2 | 450 | 900 |
| NodeMcu | Equipment | 1 | 800 | 800 |
| SensiBLE IoT | Equipment | 1 | 9000 | 9000 |
| Bluetooth HC06 Module | Equipment | 2 | 400 | 800 |
| Simblee | Equipment | 1 | 5560 | 5560 |
| MPU 6050 | Equipment | 2 | 1000 | 2000 |
| BLEduino | Equipment | 1 | 5000 | 5000 |
| ITDB02 4.3 | Equipment | 1 | 5660 | 5660 |