Overview / Usage

This analysis will help to find the person is mimicking the face or not. In this paper, we have presented a nonlinear analysis and a predictive model of voice emotion by extracting the features of a given user voice. Using the phase space plot method the data is categorized into different parameters which help to extract the feature of the voice signal by measuring the volume of the fitted ellipse on the main cluster. Using the quantifying parameter as a feature the voice signal is trained using machine learning algorithms and then Fuzzy K-Mean clustering is done to differentiate between multiple emotions. Our experimental results provide a satisfactory conclusion in this context.

Methodology / Approach

In this project, audio samples were collected from an open-source platform and classified into two types ‘Happy’ and ‘Angry’ emotions. Then using MATLAB Phase space plot method is applied. From Phase space plotting graphs insight data were obtained and observed. The insight data were about SD1, SD2, SD2/SD1, and VOL. Where SD1 and SD2 represent the major and minor axis of the ellipse and VOL represents the volume of the elliptical cluster. To discriminate the two different emotion signal we used K-mean clustering as it’s one of the best ways to differentiate the set of data.

Technologies Used

We have used MATLAB for feature extraction and Python for the prediction. The libraries used are Numpy, Matplotlib, Pandas, Kmeans from sklearn.cluster and Seaborn

Repository

https://github.com/tanmoy1999/Emotion_Voice_Analysis

Other links

• Paper Published in SSRN Journal