This research project at the Gdańsk University of Technology aims to develop a method for identifying characteristic frequency ranges for different languages. The goal is to review methods of speech frequency analysis and, based on speech recordings for various languages, verify Dr. Alfred Tomatis's theory that different languages have distinct frequency ranges for specific phonemes.
Clone the repository:
git clone https://github.com/ack2406/speech-analysis
cd speech-analysis
Install the required Python packages:
pip install -r requirements.txt
This project uses a Makefile
to streamline various tasks like downloading datasets, extracting features, and listing available options.
To download speech data and extract features for specific languages in one step:
make download_and_extract LANGUAGES="pl en" DATA_SIZE=2 FEATURES="pitch mfcc"
LANGUAGES
: List of languages to download and extract features for (e.g., "pl en"
).DATA_SIZE
: Total size of the dataset to download in GB (e.g., 2
).FEATURES
: List of features to extract (e.g., "pitch mfcc"
). If not provided, all available features will be extracted.RAW_DATA_DIR
: Directory to save the downloaded data (default: data/raw
).FEATURES_DIR
: Directory to save the extracted features (default: data/features
).ZIPS_DIR
: Temporary directory for storing downloaded zip files (default: data/zips
).To only download speech data for specific languages:
make download LANGUAGES="pl en" DATA_SIZE=2
LANGUAGES
: List of languages to download (e.g., "pl en"
).DATA_SIZE
: Total size of the dataset in GB (e.g., 2
).RAW_DATA_DIR
: Directory to save the downloaded data (default: data/raw
).ZIPS_DIR
: Temporary directory for storing downloaded zip files (default: data/zips
).To only extract features from already downloaded data:
make extract LANGUAGES="pl en" FEATURES="pitch mfcc"
LANGUAGES
: List of languages to extract features for (e.g., "pl en"
).FEATURES
: List of features to extract (e.g., "pitch mfcc"
). If not provided, all available features will be extracted.RAW_DATA_DIR
: Directory where the downloaded data is stored (default: data/raw
).FEATURES_DIR
: Directory to save the extracted features (default: data/features
).To list all available languages from Mozilla Common Voice:
make list-languages
To list all available features that can be extracted:
make list-features
data/raw
: Stores raw, downloaded audio files.data/features
: Contains extracted features (e.g., pitch, MFCC) in CSV format.data/zips
: Temporary folder for downloaded zip files.This project is licensed under the MIT License - see the LICENSE file for details.