猫咪状态识别数据集

来自米兰大学计算机系的几位猛男，和生物系、兽医系的小伙伴们一起，收集了21只猫咪的400+条语音数据，做了个猫叫声数据集。

这21只猫咪中，包括10只成年缅因猫和11只成年欧洲短毛猫，是长毛和短毛猫咪中的代表类型。
3种猫咪叫声，总长21分钟

制作这个数据集的初衷，是为了人类能更好地理解猫咪叫声的含义，为以后的人-猫沟通作准备。

因此，在制作数据集时，研究者们选择了3种让猫发出不同叫声的场景：

等待投喂：主人喂食前的猫叫声。

被放到陌生环境下隔离：将猫运送到一个陌生环境，车程不超过半小时，被隔离前会与主人先相处半小时。隔离时的猫会发出叫声。

刷牙：主人给猫刷牙时的猫叫声。

在这些场景中，每次收集猫叫的时间不超过5分钟（例如，5分钟后就会给猫喂食）。

为了让语音数据更清晰、且不影响猫的日常行动，研究者们采用了一个不到50g的小型麦克风，挂在猫脖子上。

Researchers from the Department of Computer Science at the University of Milan, collaborating with colleagues from the Department of Biology and Department of Veterinary Medicine, collected over 400 audio recordings from 21 cats and compiled a cat vocalization dataset. Among these 21 cats, there are 10 adult Maine Coons and 11 adult European Shorthairs, which are representative breeds of long-haired and short-haired cats respectively. The dataset contains three types of cat vocalizations, with a total duration of 21 minutes. The original intention of compiling this dataset is to enable humans to better understand the meaning of cat vocalizations and lay the groundwork for future human-cat communication. Therefore, when compiling the dataset, the researchers selected three scenarios that elicit distinct cat vocalizations: 1. Waiting for feeding: Cat vocalizations made by cats before being fed by their owners. 2. Isolation in a novel environment: Cats were transported to an unfamiliar environment with a car ride duration of no more than 30 minutes, and they spent 30 minutes interacting with their owners before isolation. Vocalizations were recorded while the cats were in isolation. 3. Tooth brushing: Cat vocalizations made during the process of owners brushing their cats' teeth. For each scenario, the duration of audio recording for cat vocalizations did not exceed 5 minutes (e.g., cats would be fed within 5 minutes after recording). To ensure clear audio quality without disrupting the cats' daily activities, the researchers used a small microphone weighing less than 50 grams, which was attached to the cats' necks.

搜集汇总

背景与挑战

背景概述

猫咪状态识别数据集包含21只猫咪在3种不同场景下的400+条语音数据，总时长21分钟，旨在帮助人类更好地理解猫咪叫声的含义。数据集经过剪辑处理，命名规则详细，适用于人-猫沟通研究。