Speaker-Identification-Using-Ensemble-Learning-Approach-Considering-CNN-As-Feature-Extractor

Speaker Identification Using Ensemble Learning Approach Considering CNN As Feature Extractor

Cite this paper

Dhar, S., Phukan, S., Gogoi, R., Jana, N.D. (2023). Speaker Identification Using Ensemble Learning With Deep Convolutional Features. In: Das, S., Saha, S., Coello Coello, C.A., Bansal, J.C. (eds) Advances in Data-driven Computing and Intelligent Systems. Lecture Notes in Networks and Systems, vol 653. Springer, Singapore. https://doi.org/10.1007/978-981-99-0981-0_9

Paper Link: https://link.springer.com/chapter/10.1007/978-981-99-0981-0_9

1. code file (in .ipynb format) of Audio Data Augmentation (file: Audio_data_augmentation).
2. code file (in .ipynb format) of Mel Spectrogram Generation from audio data (file: Mel_Spectrogram_Generation).

Datasets:

Input audio data are available in the given google drivelink provided below:

1. VCC2016 Dataset

https://drive.google.com/drive/folders/1LT13hS3DqBTEjwyZ5l_88Z1pcU8kYpa0?usp=sharing

This folder contains audio files of 4 speakers.

2. VCC2020 Dataset

  https://drive.google.com/drive/folders/1dy27QHZ-P7Db_1C1nHG3quL3lR1ux0Nj?usp=sharing

This folder contains Audio files of 4 speakers.

Where 2nd Letter of the folder refers the language.

• E - English
• F - Finnish
• G - German
• M - Mandarin

Where 3rd Letter of the folder refers the gender.

• F - Female
• M - Male

3. Augmented Audio Data

Due to the limited audio data of the VCC2020 Dataset, we have performed augmentation on it. Generated Audio datas are available in the given link provided below:

https://drive.google.com/drive/folders/1G3e2JFcDB0z6AETe3teJNlR3UnH4TCTS?usp=sharing

Three audio augmentation techniques are applied to generate these data:

• Pitch Scale
• Random Gain
• Invert Polarity

4. VCC2020 Merge Dataset

https://drive.google.com/drive/folders/1owS6Ml9onVOP9K_peZFGqlqszjVFlxZG?usp=sharing

We have prepared this dataset by merging data from the original VCC2020 dataset and augmented data.

5. Generated Mel Spectrogram

A) From VCC2016 Dataset

    https://drive.google.com/drive/folders/1zo_ObMLI2CXTd1mV8_62U6iwCBRo8jmY?usp=sharing

B) From VCC2020 Merge Dataset

  https://drive.google.com/drive/folders/1u2aov4Ykw3QUI50V6M9Ho9K9-OLoPzL3?usp=sharing

C) From Emotional Dataset

  https://drive.google.com/drive/folders/1lZiIDj7I7HWhFzr0Y74i-bULKQq4ta22?usp=sharing

PPT regarding the data preprocessing:

https://docs.google.com/presentation/d/1xlOfgeGpcYqbxfGoHROtlb6ngwmFzjGbNEj8__Hz4B0/edit?usp=sharing