IDMT-ISA-PUCKS

Reference

Sascha Grollmisch, David Johnson, Tobias Krüger, Judith Liebetrau

Plastic Material Classification using Neural Network based Audio Signal Analysis, in Sensor and Measurement Science International (SMSI) (submitted), Nürnberg, 2020.

Dataset Overview

The IDMT-ISA-PUCKS dataset (IIPD) was designed to simulate the challenging acoustic analysis conditions consistent with industrial manufacturing settings. The dataset contains audio recordings of multiple games of air-hockey played with pucks of different plastic materials. Data collection was performed by equipping the air hockey table with two sE8 microphones, each recording one side of the table, as seen in the image above, while a game is played. Additionally, there are recordings where no game was being played and only background noise was recorded.

We recorded the games played with different pucks at three different noise levels: Level 1 at room volume (vol_000), Level 2 with some background noise (vol_050 = 70 CBR) and Level 3 at loud background noise (vol_100 = 80 CBR). The background noise was played over four speakers in equal distances around the table and contains human voices.

The following materials were used for the four pucks:

  • Puck_A is the original factory puck (material unknown)
  • Puck_E from the 3D printer (material: ABS, print process: FDM)
  • Puck_G from the 3D printer (material: PA2200, print process: SLS)
  • Puck_I from the 3D printer (material: PA12, print process: MJF)

For each noise level and puck material, five three-minute games were played with different pucks of the specified material. Further, each game was played with different sets of players. The recordings were made via two sE8 microphones placed in the middle of the air-hockey table (about 10 cm above the surface).

Dataset total duration:  260 minutes (1 min per file)

  • # Files for puck_A: 45
  • # Files for puck_E: 45
  • # Files for puck_G: 45
  • # Files for puck_I: 45
  • # Files for no_puck: 45
  • # Total WAV Files: 260
  • Sampling rate: 44.1KHz
  • Resolution: 32-bit
  • Stereo audio