Speech emotion detection using Catboost model trained by labelling LPC, MFCC, Pitch features of speech from a emotion classified speech dataset -- 2

I want to develop code to do emotion detection on speech by training a model by pre-processes the test voice signal, extract selective MFCC, LPC, pitch, and voice quality features,
then classify the speech using Catboost.
Step 1. Read each file with different sentiments and languages
Step 2. Filter the voices. We removed blanks
Step 3. Extract LPC and MFCC features. Calculate Pitch. Club them all as features
Step 4. Features satisfying a threshold are retained
Step 5. Assign labels to features as per the sentiment in the voice
Step 6. Design a Catboost model having relevant parameters
Step 7. Using features and labels train the model
Step 8. To predict the sentiment
Take a new voice for output
Repeat steps 2 through 4
Run the model with the new features to predict the sentiment
Use Catboost to extract the categorical features and to predict the sentiments from voices.


A detailed description is there in the attached research paper.
All these algortithms have to be implemented exactly as described in the research paper.

Training datasets
1. RAVDESS dataset
https://www.kaggle.com/uwrfkaggler/ravdess-emotional-speech-audio
2. TESS dataset
https://www.kaggle.com/ejlok1/toronto-emotional-speech-set-tess
3. Emo-DB Dataset
https://www.kaggle.com/piyushagni5/berlin-database-of-emotional-speech-emodb


Deliverables:-
A simple web page should be created where:-
1. A new dataset can be uploaded and the Catboost model can be trained using that dataset.
2. A speech audio is uploaded and the model predicts the emotion.
3. Full detailed documentation explaining every line of code.