Joint Discrete and Continuous Emotion Prediction Using Ensemble and End-to-End Approaches

AlBadawy, Ehab A.; Kim, Yelin

doi:10.1145/3242969.3242972

Cited by 8 publications

(6 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These performances were obtained using an optimizable ensemble regressor. Our performances are better than those from the literature [11][12][13][14][15] (see Section 2), who performed more complex processing and feature extraction. In Table 1, the validation performances were evaluated by performing 5-fold cross validation across the training data.…”

Section: Discussion Of Resultsmentioning

confidence: 60%

“…They achieved a CCC of 0.413 and 0.527 on arousal and valence predictions, respectively. Albadawy et al [15] used the visual features provided by AVEC 2015, which included appearance (LGBP-TOP) and geometric (Euclidean distances between 49 facial landmarks) features. For arousal and valence predictions, they proposed a joint modelling strategy using a deep BiLSTM for ensemble and end-to-end models.…”

Section: Literature Overviewmentioning

confidence: 99%

See 1 more Smart Citation

Optimizable Ensemble Regression for Arousal and Valence Predictions from Visual Features

Joudeh,

Cretu,

Bouchard

2023

Ecsa 2023

View full text Add to dashboard Cite

show abstract

Section: Discussion Of Resultsmentioning

confidence: 60%

Section: Literature Overviewmentioning

confidence: 99%

Optimizable Ensemble Regression for Arousal and Valence Predictions from Visual Features

Joudeh,

Cretu,

Bouchard

2023

Ecsa 2023

View full text Add to dashboard Cite

show abstract

“…log Mel filterbank LSTM [8] LSTM [28] 0.527(1) 0.504(2) BottleNeck LSTM [29] 0.533 0.466 eGeMAPS + BoAW RVM [10] 0.494 0.507 eGeMAPS ARX 0.540 0.502 It can be observed that ARX achieves comparable performances to the much more complex non-linear LSTM back-ends on all three databases. It is also worth noting that it outperforms the best LSTM system on the SEWA database, with 1.3% relative improvement for arousal.…”

Section: Resultsmentioning

confidence: 98%

“…While a significant number of regression modelling techniques have been proposed for emotion predictions, the most widely adopted models are Support Vector Regression [6] owing to its robustness, and Long-short term memory recurrent neural network (LSTM-RNN) [7][8][9] which captures the temporal dynamics of emotion. In addition, relevance vector machines (RVM) [10] and Gaussian mixture regression (GMR) [11] have also shown good performance in emotion prediction tasks.…”

Section: Introductionmentioning

confidence: 99%

Speech Based Emotion Prediction: Can a Linear Model Work?

Ouyang¹,

Dang

Sethu

et al. 2019

Interspeech 2019

View full text Add to dashboard Cite

Speech based continuous emotion prediction systems have predominantly been based on complex non-linear back-ends, with an increasing attention on long-short term memory recurrent neural networks. While this has led to accurate predictions, complex models may suffer from issues with interpretability, model selection and overfitting. In this paper, we demonstrate that a linear model can capture most of the relationship between speech features and emotion labels in the continuous arousal-valence space. Specifically, an autoregressive exogenous model (ARX) is shown to be an effective backend. This approach is validated on three commonly used databases, namely RECOLA, SEWA and USC CreativeIT, and shown to be comparable in terms of performance to state-of-the-art LSTM systems. More importantly, this approach allows for the use of well-established linear system theory to aid with model interpretability.

show abstract

“…Khorram et al [66] proposed two architectures for continuous emotions recognition-dilated CNN with a varying dilation factor for different layers and downsampling/upsampling CNN-with different ways of modelling long-term dependencies. AlBadawy and Kim [68] further improved the accuracy of valence with joint modelling of the discrete and continuous emotion labels. Table 4 summarises the top performances of the continuous SER systems tested on the RECOLA dataset.…”

Section: Temporal Variations Modellingmentioning

confidence: 99%

A Review on Speech Emotion Recognition Using Deep Learning and Attention Mechanism

et al. 2021

View full text Add to dashboard Cite

Emotions are an integral part of human interactions and are significant factors in determining user satisfaction or customer opinion. speech emotion recognition (SER) modules also play an important role in the development of human–computer interaction (HCI) applications. A tremendous number of SER systems have been developed over the last decades. Attention-based deep neural networks (DNNs) have been shown as suitable tools for mining information that is unevenly time distributed in multimedia content. The attention mechanism has been recently incorporated in DNN architectures to emphasise also emotional salient information. This paper provides a review of the recent development in SER and also examines the impact of various attention mechanisms on SER performance. Overall comparison of the system accuracies is performed on a widely used IEMOCAP benchmark database.

show abstract

Joint Discrete and Continuous Emotion Prediction Using Ensemble and End-to-End Approaches

Cited by 8 publications

References 34 publications

Optimizable Ensemble Regression for Arousal and Valence Predictions from Visual Features

Optimizable Ensemble Regression for Arousal and Valence Predictions from Visual Features

Speech Based Emotion Prediction: Can a Linear Model Work?

A Review on Speech Emotion Recognition Using Deep Learning and Attention Mechanism

Contact Info

Product

Resources

About