Speech Emotion Recognition Adaptation Benchmark (SERAB)

Speech Emotion Recognition Adaptation Benchmark (SERAB) 是一个多语言的语音情感识别基准，由洛桑联邦理工学院等机构创建。该数据集包含九个子数据集，覆盖六种语言，旨在评估不同语音情感识别方法的性能和泛化能力。数据集大小、情感类别数量各异，适用于评估预训练深度神经网络特征提取器的泛化能力。SERAB 的应用领域广泛，包括人机交互、医疗健康和公共安全等，旨在解决自动准确分类语音中情感的挑战。Speech Emotion Recognition Adaptation Benchmark (SERAB) is a multilingual speech emotion recognition benchmark developed by institutions including the École Polytechnique Fédérale de Lausanne (EPFL). This dataset consists of nine sub-datasets spanning six languages, designed to evaluate the performance and generalization capability of diverse speech emotion recognition methods. The sub-datasets vary in scale and number of emotion categories, making them suitable for assessing the generalization ability of pre-trained deep neural network-based feature extractors. SERAB has broad application domains, including human-computer interaction, healthcare, public safety and more, aiming to address the challenge of automatically and accurately classifying emotions in speech.

搜集汇总

背景与挑战

背景概述

SERAB是一个多语言语音情感识别基准数据集，包含九个子数据集，覆盖六种语言，用于评估语音情感识别方法的性能和泛化能力。该数据集适用于预训练深度神经网络特征提取器的测试，应用领域包括人机交互、医疗健康等，旨在解决语音情感自动分类的挑战。