---
license: cc-by-nc-4.0
---
# VibroTemp092024
## A vibroacoustic temperature dataset

For research details, check out our paper: [Temperature Prediction extracted from Tissue-Needle Vibroacoustic Signals: Initial Findings](https://doi.org/10.36227/techrxiv.173386402.21219983/v1).

The dataset consists of three folders:
- annotations,
- audio,
- snippets

Below are the descriptions of the files in each of the folders.

### Annotations

Each annotation file is a JSON file with the following keys:
1. audio_file - name of the audio file recorded during the measurement
2. audio_annotations - manual annotation, based off the video recordings, of relevant events (in samples):
  - 1 -> the moment the needle tip enters the foam
  - 2 -> the stop of the needle forward movement in the foam
  - 3 & 4 -> not relevant
3. video_file - name of the video file recorded during the audio measurement - not relevant
4. video_annotations - same as audio_annotations, but the frame of the event is given - not relevant

### Audio

The audio files collected during the vibroacoustic measurements. There are two channels - the first corresponds to the needle microphone, and the second corresponds to the environment microphone, which was used for noise cancellation.

### Snippets

The cut audio files which were converted to Mel spectrograms and used as input to the deep learning model.