jcvasquezc/DisVoice

feature extraction from speech signals

What it solves

DisVoice provides a standardized way to extract acoustic and linguistic features from speech files to help identify paralinguistic aspects of voice. This is primarily used to recognize emotions or detect communication impairments caused by speech disorders, such as Parkinson's, Huntington's, larynx cancer, or cleft-lip and palate, as well as mood disorders like depression.

How it works

The framework computes a wide variety of speech features from both sustained vowels and continuous speech. It utilizes several extraction strategies:

  • Glottal, Phonation, Articulation, and Prosody: Traditional acoustic analysis.
  • Phonological: Analysis of speech sounds.
  • Representation Learning: Using autoencoders to learn feature representations from the audio data.

It integrates with external tools like Praat and Kaldi for the underlying signal processing and phonological analysis.

Who it’s for

Researchers and clinicians focusing on speech pathology, medical diagnostics via audio, and emotion recognition.

Highlights

  • Supports extraction of glottal, phonation, articulation, prosody, and phonological features.
  • Uses autoencoders for representation learning.
  • Capable of analyzing both sustained vowels and continuous speech.
  • Designed for the classification of pathological speech and mood-related patterns.

Related

  • Project
    audeering/opensmile-python

    A Python interface for openSMILE that enables the extraction of standardized audio features for machine learning research.

  • Project
    QwenAudio/SenseVoice

    A high-accuracy, low-latency speech foundation model that provides multilingual speech recognition, emotion detection, and audio event identification.

  • Project
    JeremyCCHsu/Python-Wrapper-for-World-Vocoder

    A Python wrapper for the WORLD Vocoder that allows for high-quality speech analysis and resynthesis by decomposing audio into pitch, spectral envelope, and aperiodicity.

  • Project
    filippogiruzzi/voice_activity_detection

    A deep learning-based Voice Activity Detection system using a 1D-ResNet and MFCC features to classify audio signals as speech or noise.

  • Project
    cmusphinx/pocketsphinx

    PocketSphinx is an open‑source, offline speech‑to‑text engine (C library with Python bindings) that runs efficiently on low‑resource devices. It provides a command‑line tool and APIs for recognizing or aligning audio, outputting results as JSON.