Pet Vocalization Emotion Annotation Dataset

This dataset contains audio recordings of cat and dog vocalizations, along with corresponding emotion and behavioral state labels. It is suitable for research and development in pet sound classification, emotion label prediction, animal sound event recognition, and interactive pet‑related products. The audio has been segmented into short clips, facilitating sample‑based retrieval, processing, and model training.

The dataset includes the following components:

• Animal Category: Includes cats and dogs; each entry specifies the corresponding animal type.
• Emotion State Labels: Sample categories cover happiness, excitement, fear, anxiety, warning, anticipation, submission, enjoyment, obedience, pleading, and more; a single audio clip may be associated with one or multiple labels.
• Chinese–English Annotations: Provides both Chinese and English labels to support multilingual search and label mapping.
• Audio Metadata: Includes fields such as filename, audio format, duration, sampling rate, and number of channels; some entries also provide bitrate, useful for sample filtering and audio preprocessing.

The audio files are provided in MP3 format, with segment durations ranging from approximately 5 to 20 seconds. The samples include two sampling rates—44.1 kHz and 8 kHz—and all recordings are stereo. Annotations are structured as JSON objects, linked to audio files via filenames; specific parameters are defined for each individual record.

A downloadable demo is available for inspection: json.txt