GoodSounds

GoodSounds 数据集包含大约 28 小时的单音和音阶录音，由 15 位不同的专业音乐家演奏，他们都拥有音乐学位并具有一定的教学专业知识。使用一个或最多 4 个不同的麦克风录制了 12 种不同的乐器（长笛、大提琴、单簧管、小号、小提琴、中音萨克斯、中音萨克斯、男中音萨克斯、高音萨克斯、双簧管、短笛和贝司）。对于所有乐器，乐器中的整套可演奏半音被记录多次，具有不同的音调特征。每个音符都被录制到一个单独的 48kHz 和 32 位的单声道音频文件中。提供丰富的录音注释，包括录音环境的详细信息和声音的音质评级（“好声音”、“坏”、“音阶好”、“音阶坏”）。The GoodSounds dataset contains approximately 28 hours of monophonic and scale recordings, performed by 15 distinct professional musicians, all of whom hold music degrees and possess relevant teaching expertise. A total of 12 distinct musical instruments (flute, cello, clarinet, trumpet, violin, alto saxophone, alto saxophone, baritone saxophone, soprano saxophone, oboe, piccolo, and double bass) were recorded using one or up to four different microphones. For every instrument, the complete set of playable chromatic notes within its range was recorded multiple times with varying timbral characteristics. Each individual note was captured as a separate monophonic audio file with a sampling rate of 48 kHz and a bit depth of 32 bits. Comprehensive recording annotations are provided, including detailed information about the recording environment and sound quality ratings: "good sound", "bad", "good scale", "bad scale".

搜集汇总

背景与挑战

背景概述

GoodSounds 是一个音频数据集，包含约28小时由15位专业音乐家演奏12种乐器的单音和音阶录音，采用48kHz和32位单声道格式，并提供详细的音质评级注释（如'好声音'、'坏'），适用于音频识别和音乐质量判定任务。