griko/height_reg_svr_ecapa_voxceleb

Height Estimation Model

This model combines the SpeechBrain ECAPA-TDNN speaker embedding model with an SVR regressor to predict speaker height from audio input. The model was trained on the VoxCeleb2 and evaluated on the VoxCeleb2 and TIMIT datasets.

Model Details

• Input: Audio file (will be converted to 16kHz, mono, single channel)
• Output: Predicted height in centimeters (continuous value)
• Speaker embedding: 192-dimensional ECAPA-TDNN embedding from SpeechBrain
• Regressor: Support Vector Regression optimized through Optuna
• Performance:
  • VoxCeleb2 test set: 6.01 cm Mean Absolute Error (MAE)
  • TIMIT test set: 6.02 cm Mean Absolute Error (MAE)

Training Data

The model was trained on height enriched VoxCeleb2 dataset (for details read the paper):

• Audio preprocessing:
  • Converted to WAV format, single channel, 16kHz sampling rate, 256 kp/s bitrate
  • Applied SileroVAD for voice activity detection, taking the first voiced segment

Installation

You can install the package directly from GitHub:

pip install git+https://github.com/griko/voice-height-regression.git

Usage

from voice_height_regressor import HeightRegressionPipeline

# Load the pipeline
regressor = HeightRegressionPipeline.from_pretrained(
    "griko/height_reg_svr_ecapa_voxceleb"
)

# Single file prediction
result = regressor("path/to/audio.wav")
print(f"Predicted height: {result[0]:.1f} cm")

# Batch prediction
results = regressor(["audio1.wav", "audio2.wav"])
print(f"Predicted heights: {[f'{h:.1f}' for h in results]} cm")

Limitations

• Model was trained on celebrity voices from YouTube interviews
• Performance may vary on different audio qualities or recording conditions
• Height predictions are estimates and should not be used for medical or legal purposes

Citation

If you use this model in your research, please cite:

@misc{koushnir2025vanpyvoiceanalysisframework,
      title={VANPY: Voice Analysis Framework}, 
      author={Gregory Koushnir and Michael Fire and Galit Fuhrmann Alpert and Dima Kagan},
      year={2025},
      eprint={2502.17579},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2502.17579}, 
}