ASPEN | Auditory Stimulus for Psychophysical ExperimeNt

ASPEN is a full python toolkit to generate Auditory Stimulus for Psychophysical ExperimeNt.

Docs | Toolkit Overview | Installation & Usage | How To Use

Purpose

• This project aims to
  • create any auditory signal to examine human perception for research
  • mainly focus on stimulus generation itself, not a platform for psychophysical experiment.

Key Features

• Support multiple auditory stimulus
  • locally time-reversed speech
  • continuity illusion
  • auditory streaming
  • noise-vocoded speech
  • modulation-filtered speech
  • verbal transformation effect
  • iterated rippled noise
• Support multiple fundamental signals
  • pure tone
  • harmonic complex tone
  • amplitude modulated tone
  • frequency modulated tone
  • click train pitch
  • colored noise (white / brown / pink / blue / violet)
  • filtered noise (lowpass / highpass / bandpass / bandstop)
• Support multiple visualization
  • waveform
  • spectrum
  • spectrogram
  • modulation power spectrum

Installation

git clone https://github.com/ashi-ta/aspen.git
cd aspen
pip install -e .

Usage

• there are some example scripts and configs
  • egs/run_*.sh are example bash scripts
  • egs/conf contain example configuration files

$ cd egs

# generate auditory streaming signals under some conditions
$ ./run_auditory_streaming.sh

• More details in docs.

References

The paper utilized this toolkit is to be published in INTERSPEECH 2021.

Takanori Ashihara, Takafumi Moriya, Makio Kashino, "Investigating the Impact of Spectral and Temporal Degradation on End-to-End Automatic Speech Recognition Performance", In Proc. INTERSPEECH, 2021