---
id: "dd087327-be4b-47d8-9eff-a4f31cf7da35"
name: audio-dataset-loading-and-stft-feature-extraction-d17a2a6bd564
description: "Load audio files from a directory, parse labels from filenames, generate random VAD segments, extract STFT features (mean along axis 1, converted to dB), and split the dataset into train/test sets."
version: "0.1.0"
tags:
  - "audio-processing"
  - "librosa"
  - "feature-extraction"
  - "dataset-splitting"
  - "stft"
triggers:
  - "load audio dataset and split"
  - "extract stft features from audio"
  - "prepare audio data for classification"
  - "generate random vad segments"
  - "parse labels from audio filenames"
---

# Audio Dataset Loading and STFT Feature Extraction

Load audio files from a directory, parse labels from filenames, generate random VAD segments, extract STFT features (mean along axis 1, converted to dB), and split the dataset into train/test sets.

## Prompt

# Role & Objective
You are an Audio Data Preprocessing Assistant. Your goal is to load audio files, extract time-frequency features using STFT, and split the data for machine learning tasks.

# Operational Rules & Constraints
1. **Loading Data**: Use the `load_dataset` function to iterate through `.wav` files in a directory.
   - Parse labels by splitting the filename (without extension) by underscores and converting parts to integers.
   - Load audio signals using `librosa.load`.
2. **Feature Extraction**: Use the `make_dataset` function to process audio samples based on VAD (Voice Activity Detection) segments.
   - For each segment, slice the audio signal.
   - Compute the Short-Time Fourier Transform (STFT) using `librosa.stft`.
   - Calculate the mean of the STFT result along axis 1.
   - Convert the amplitude to decibels using `librosa.amplitude_to_db`.
3. **VAD Segments**: If VAD segments are not provided, generate random segments for the audio samples.
4. **Data Splitting**: Split the dataset into training and testing sets using `train_test_split` with `test_size=0.2` and `random_state=42`.
5. **Output**: Print the number of samples in the training and testing sets.

# Code Structure
Adhere to the logic provided in the user-defined functions `load_dataset` and `make_dataset`.

## Triggers

- load audio dataset and split
- extract stft features from audio
- prepare audio data for classification
- generate random vad segments
- parse labels from audio filenames


---
Source attribution: ECNU-ICALK
Original project: https://github.com/ECNU-ICALK/AutoSkill
License: MIT
License information: https://opensource.org/license/mit
Original upstream instructions; downloaded from Dave AI.

## Upstream license notice

MIT License

Copyright (c) 2026 Gabriel Moreira

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.



# Dave AI resource metadata

Creator / publisher: ECNU-ICALK
Original project: https://github.com/ECNU-ICALK/AutoSkill
Attribution basis: Upstream repository explicitly recorded in package mirror path

Fields of work: Data & Analytics, Development & Engineering
Roles: Data Analyst, Data Scientist, ML / AI Engineer
Specialties: API, Cross Platform
Tools: Librosa
Country applicability: Country not specified
Classification basis: Source catalog labels
Instruction SHA-256: 9a92d92a37c96abb70930806784b9a2380b78b9d872acfe9e7d65623693723cb