Skip to content

This repository contains the approach and code used by me, in implementing this project. This is also my final year project of my engineering curriculum.

License

Notifications You must be signed in to change notification settings

bumblebee26/TV-audience-measurement-and-Sentiment-Analysis

Repository files navigation

TV audience measurement and Sentiment Analysis

A smart system which measures the television audience as well as provides valuable insights about the viewers to the channel providers.

Features

  1. Check the authenticity of the channels
  2. Total viewers watching the channel/program
  3. Viewer’s demographic information like
  • Total members in a family
  • Gender
  • Age
  • Emotion

Work Flow of the product

Flow diagram

Authentication of Channels via digital watermarking

Digital Video Watermarking the channel provider content with hidden watermark and extracting it to check its authenticity and report malicious activities

Watermarking Demo

Youtube Demo

Checking the Robustness of the watermark with attacks

Gaussian Noise | PSNR = 65.88 Salt and Pepper Noise | PSNR = 69.4819
Cropping | PSNR = 37.3145 Sharpening | PSNR = 71.3505
Histogram Equalization | PSNR = 49.0789 Intensity Adjustment | PSNR = 68.2943

Viewer's Demographic Data Collection

The smart meter consists of a camera which captures the viewers in front of the television. This captured frames are passed through the recognition software that detects gender, age and emotion of the viewer.

Facial Emotion Recognition Dataset

FER dataset training on Kaggle

Age, Gender and Emotion Recognition Demo

Youtube Demo

Collected Viewer's Data

Data Visualization

A website running on the local server, where the measurement authority can login and check the visualizations.

Other Authors

Yogesh Deshpande

Akshay Bhogan