Documentation Software

Software Architecture

Overview of the software infrastructure, data processing pipelines, and utilities used in the TAPS dataset.

Data Processing Pipeline

Signal Processing

  • • 5th-order Butterworth high-pass filter (50Hz cutoff)
  • • DC offset removal
  • • Signal normalization
  • • Timing synchronization between devices

Data Formats

  • • WAV files (32-bit)
  • • Throat mic: 8kHz sampling rate
  • • Acoustic mic: 16kHz sampling rate
  • • Standard file naming convention

Software Tools

Data Collection Software

Custom-built software for simultaneous recording from throat and acoustic microphones.

Features

  • • Real-time monitoring
  • • Automatic file naming and organization
  • • Recording quality validation
  • • Session management

Data Processing Tools

Suite of tools for data preprocessing, validation, and analysis.

Capabilities

  • • Automatic signal alignment
  • • Noise reduction
  • • Quality metrics calculation
  • • Batch processing support

Data Structure

File Organization

dataset/
├── train/
│   ├── speaker_001/
│   │   ├── p01_u00_mic.wav
│   │   ├── p01_u00_acc.wav
│   │   └── ...
│   └── ...
├── eval/
│   ├── speaker_051/
│   │   ├── p51_u00_mic.wav
│   │   ├── p51_u00_acc.wav
│   │   └── ...
│   └── ...
└── metadata/
    ├── transcripts.txt
    └── speaker_info.json