Skip to main content

Vocal Fry and other Metrics

Project description

Resemble Speech Tools

This package contains collated tools for quantifying issues in speech clips.

The repo currently has support for detecting vocal fry (H2H1 metric proposed by Kane-Drugman):

Usage:

from speechtools.features import get_creak_features

get_creak_features(waveform, fs, f0_min=20, f0_max=500, window=1600, hop=160, use_fixed_windows=False, return_frames=False

Special Parameters
  • use_fixed_frames: This needs to be True if user wants to override internally used window and hop sizes. If False, the default window and hop sizes specified by the KaneDrugman papers are used for analysis.
  • return_frames: When False the resulting metric is interpolated to be of waveform length.

Miscellaneous

There are also a few private functions bundled in for signal processing under speechtools.audio_utils:

  • Splitting up the signal into chunks of win_length strided by hop_size
  • Obtaining frame level LPC residuals

Papers:

  1. Kane, J., Drugman, T., Gobl, C., (2013) "Improved automatic detection of creak", 27(4), pp. 1028-1047, Computer Speech and Language.
  2. Drugman, T., Kane, J., Gobl, C. (2012) "Resonator-based creaky voice detection", Proceedings of Interspeech.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

speech-analysis-tools-0.2.tar.gz (5.4 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page