MSP-IMPROV: An Acted Corpus of Dyadic Interactions to Study Emotion Perception

By Carlos Busso, Srinivas Parthasarathy, Alec Burmania, Mohammed Abdel-Wahab, N. Sadoughi, Emily Mower Provost. 2017

401 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Experimental and Cognitive Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

1. Kun Zhou
2. Lele Chen
3. Haizhou Li
4. Chenliang Xu
5. Zhaoxi Mu
6. Carlos Busso
7. Xi Li
8. Xinyu Yang

In Experimental and Cognitive Psychology

1. Qin Jin
2. Jinming Zhao
3. Reza Lotfian
4. Rui Liu
5. Kun Zhou
6. Lele Chen
7. Haizhou Li
8. Chenliang Xu

Citations in the corpus

• The Emotional Voices Database: Towards Controlling the Emotion Dimension in Voice Generation Systems (2018, background): “IMPROV contains 6 sessions from 12 actors resulting in 9 hours of audiovisual data.”
• Building Naturalistic Emotionally Balanced Speech Corpus by Retrieving Emotional Speech from Existing Podcast Recordings (2019, methodology): “For this analysis, we combine the IEMOCAP [10] and MSP-IMPROV [19] databases, creating two partitions, one for training (50 percent) and one for testing (50 percent) the models.”
• Emotional Voice Conversion: Theory, Databases and ESD (2021, methodology): “Similar to IEMOCAP, MSP-IMPROV (Busso et al., 2016) consists of 6 sessions and 9 h of audio–visual data from 12 actors.”
• What comprises a good talking-head video generation?: A Survey and Benchmark (2020, background): “2018) and MSP-IMPROV dataset (Busso et al. 2016) involve creating stimulus with conflicting emotional content conveyed through speech and facial expression.”
• Missing Modality Imagination Network for Emotion Recognition with Uncertain Missing Modalities (2021, methodology): “The bottom block in Table 5 shows the performance comparison on the MSP-IMPROV dataset.”
• Review of end-to-end speech synthesis technology based on deep learning (2021, background): “None of the abovementioned corpora contains explicit emotional information, and most of the existing emotional corpora cannot be effectively used to train the emotional TTS model based on deep learning, because these data sets contain a small number of sentences, such as RAVDESS [128], CREMA-D [23],”
• Recent Advances and Trends in Multimodal Deep Learning: A Review (2021, methodology): “MSP-Improve [19] database was compiled to record naturalistic emotions from improvised circumstances.”