The Music-Mouv' Dataset

This page contains the data collected in the context of the Music-Mouv' project (2021-2023), which was a collaboration between two groups of the Université de Lorraine:

• The BIRD group (Building artificial Intelligence between trust, Responsibility and Decision) of the Loria (Lorraine Laboratory of Research in Computer Science and its Applications)
• The CEMA axis (Cognition-EMotion-Action) of the 2LPN lab (Lorraine Laboratory in Psychology and Neuroscience)

The goal of the project was to facilitate gait initiation (more precisely the anticipatory postural adjustments, which are problematic to the elderly, to people with Parkinson's disease, etc.) by inducing relevant emotions through music listening. We collected subjective, physiological and biomechanical data using questionnaires, wristbands and shoe insoles equipped with sensors, to predict future emotions induced by music listening and to study the effects of emotions on gait initiation.

If you have any questions or comments, feel free to e-mail the project's principal investigator: bonnin [at] loria [dot] fr

The dataset

Download (282 MB)

We provide the data in JSON format. There are 35 JSON files in total, one per participant. Each file has the following structure:

{
  "id": 10,
  ...,
  "test_i": {
    "insole_data": {
      "timestamp": 1650620059000,
      "duration_s": "50s",
      "pressure_g/cm²": {
        "time_s": [ 0.0, 0.01, ... ],
        "left_foot": [ 28.5184, 28.5353, ... ],
        "right_foot": [ 26.1726, 26.1955, ... ]
      }
    },
    "watch_data": {
      "blood_volume_pulse": {
        "value": [ -3.55, -6.28, ... ],
        "time_s": [ 837.0, 837.015625, ... ]
      },
      "electro_dermal_activity": {
        "value": [ 7.586999, 7.566499, ... ],
        "time_s": [ 837.0, 837.25, ... ]
      },
      "heart_rate": {
        "value": [ 102.97, 103.03, ... ],
        "time_s": [ 827, 828, ... ]
      },
      "temperature": {
        "value": [ 33.16, 33.16, ... ],
        "time_s": [ 837.0, 837.25, ... ]
      }
    },
    "music_data": {
      "spotify_data": { ... },
      "music_style": "house"
    },
    "interview": {
      "test_number": "1",
      "appreciation": true,
      "familiarity": true,
      "emotional_perception": {
        "emotional_perception_original": [ "dynamisme", "entrain" ],
        "emotional_perception_translated": [ "dynamic", "spirited" ]
      },
      "standardized_esthetic_emotion": "Joyful Activation",
      "standardized_esthetic_classes": "Vitality",
      "emotion_sam_pleasure": 4.0,
      "emotion_sam_arousal": 3.0,
      "other_feedback": {
        "other_feedback_original": "entrainé, vouloir avancer",
        "other_feedback_translated": "carried away, wants to move forward"
      },
      "effect_on_walking": {
        "effect_on_walking_original": "plus dynamique, premier pas plus rapide",
        "effect_on_walking_translated": "more dynamic, faster first step"
      }
    }
  },
  ...
}

Each "test_i" variable corresponds to one of the participant's trials and contains the following variables:

• insole_data: plantar pressure of each foot collected with Technoconcept Dynafoot 2 insoles
• watch_data: blood volume pulse, electro dermal activity, hear rate and temperature collected with Empatica E4 wristbands
• music_data: characteristics of the track played during the trial, extracted using Spotify's API
• interview: subjective answers to the Self-Assessment Manikin (SAM) questionnaire and to a semi-directed interview to collect emotion categories from the Geneva Emotional Music Scale (GEMS).

References

If you use this data, please cite the following paper:

Méhania Doumbia, Maxime Renard, Laure Coudrat, and Geoffray Bonnin. 2023. Characterizing the Emotional Context Induced by Music Listening and its Effects on Gait Initiation: Exploiting Physiological and Biomechanical Data. In Adjunct Proceedings of the 31st ACM Conference on User Modeling, Adaptation and Personalization (UMAP '23 Adjunct). Association for Computing Machinery, New York, NY, USA, 182–186. https://doi.org/10.1145/3563359.3596982

MuRS@RecSys'24

We also published a paper in the Music Recommender Systems Workshop at Recsys, which is based on an extended version of the dataset with physiological features extracted from the raw physiological data. We provide the new data as a set of CSV files:

• eda_features.csv: A few peak-based Electrodermal Activity features.
• hrv_features.csv: A few Heart Rate Variance related features.
• music_features.csv: The music features provided by the API of Spotify.
• familiarities.csv: Binary music familiarity ratings (familiar/unfamiliar).
• emotions.csv: The GEMS categories of induced emotions.

We also provide the Python code we used in our experiments.

Download