Timbre Classification experiments

This repository contains datasets and scripts for timbre classification experiments conducted as part the Ph.D. thesis. Two datasets were used. The first one concentrates on drum/percussion sounds while the other generalises to orchestral sounds. See the relevant iPython notebooks to re-run experiments.
The orchestral sample is quite large, there is a script that pulls N number samples randomly in the folder, for performing smaller analyses. Each episode directory contains word-level and segment-level information of the whole episode and also parallel samples extracted under segments_eng and segments_spa subdirectories. Each sample is stored as an WAV audio file, text file and a CSV file containing word timing information and word-level paralinguistic and prosodic features.

本仓库包含用于音色分类（timbre classification）实验的数据集与脚本，这些实验是博士学位论文研究的一部分。实验采用两个数据集：其一聚焦鼓/打击乐器音色，其二则覆盖管弦乐音色。如需复现实验，可参考相关iPython笔记本。管弦乐样本规模较大，仓库中提供了一个脚本，可从文件夹内随机抽取N个样本以开展小规模分析。每个片段目录包含该片段的词级与段级信息，以及在segments_eng和segments_spa子目录下提取的平行样本。每个样本以WAV音频文件、文本文件及CSV文件形式存储；CSV文件包含词时序信息、词级副语言特征（paralinguistic features）与韵律特征（prosodic features）。