Project: Vietnamese Accent Recognition using Deep Learning
The project runs on Google Colab and includes the following sections:
- Data preprocessing and feature extraction (MFCC - Mel Frequency Cepstral Coefficients)
- Utilizes two types of models to train and predict: CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network)
- The dataset used in this project is the Vietnamese Common Voice provided by Mozilla. Mozilla started this project to create a free database for developers to build various voice recognition software. As of the completion of this project, Mozilla has developed datasets in multiple languages, including Vietnamese. More details can be found at: Mozilla Common Voice.
- This project uses a portion of the Common Voice Corpus 9.0 dataset (updated on 27/04/2022). The downloaded data includes voice recordings of many people from different ages and regions, along with an Excel file containing the corresponding sentences for each voice recording.
- vietnamese_accent_recognition.ipynb: The main Jupyter Notebook file of the project.
- audio_record_details.xlsx: Excel file containing the audio recording information.
- audio_records: Folder containing the audio recordings.
The project was completed by a group of two members: