This week was focused on gathering data, picking algorithms, and preprocessing/getting features of data
Next week will be finishing gathering important features from data and starting to train.
SOW and Gantt Chart:
The original web application was going to communicate with the RPi and receive textual journal entries of the user. Since we will no longer be making the physical product, the tasks related to setting up communication between the Raspberry Pi and web application had to be replaced.
Therefore I will work on speech tone analysis to complement Patrick’s work on facial emotion analysis and Vinay’s sentiment/text analysis. Speech tone analysis uses the voice waveforms to train and classify emotions. Waveforms contain information like pitch and formants that can be used as features. The classifier used to train will be an SVM because the size of our dataset is relatively small and there are not many features that need to be extracted.
I will be using Praat (http://www.fon.hum.uva.nl/praat/) voice analysis software to extract voice waves and measure values such as pitch and formants from sound recordings. The datasets I will use is the Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) which can be found here: https://www.kaggle.com/uwrfkaggler/ravdess-emotional-speech-audio?fbclid=IwAR2qFqomaS_N32Oke3Lhbil-wKsI35LZFrY55tMveIkT1ib2nCZxgCiywdo and the Toronto emotional speech set (found here: https://tspace.library.utoronto.ca/handle/1807/24487).
Additional Info:
The minimum analyzer I would like to have working is the binary classier for high/low arousal. Because emotion categories can be categorized into high/low arousal (happy, angry, surprise being high, sad, disgust, fear being low), a binary classier may help be the final decider if the facial emotion and text analysis do not match up.
Afterwards, I would also like to be able to write an analyzer to classify the voices into the six different categories, using an SVM like I mentioned above. Papers I found recommended a linear kernel SVM.
Gantt chart below. Schedule is tight, hopefully will be able to finish earlier.
0 Comments