Two-stage dimensional emotion recognition by fusing predictions of acoustic and text networks using SVM

Atmaja, Bagus Tris; Akagi, Masato

doi:10.1016/j.specom.2020.11.003

Cited by 37 publications

(17 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If the methods of feature selection and weight calculation are different, the document will be converted into different styles of improved SVM model. According to the theory of the improved SVM algorithm, the samples of nonsupport vectors are bound to be classified correctly in case, so these samples are classified [24]. Second, the theoretical method is used to analyze each support vector in the emotional model of online teaching and calculate it, so as to dig out the potential relationship between each support vector and other support vectors, which can be used as the characteristics of this support vector in the situation.…”

Section: Improved Svm Model and Algorithmmentioning

confidence: 99%

Analysis of Affective Factors and Optimization Strategies of Emotion in Online Teaching Based on Improved SVM Model

Huwan²

2022

Security and Communication Networks

View full text Add to dashboard Cite

Online teaching has the advantages of not being limited by location and space but it also has some shortcomings. The lack of face-to-face real-time interaction between teachers and students will affect some students’ learning mood. The improved support vector machine (SVM) model is a simple model based on linear algebra, which can convert text data into structured data that can be processed by a computer and then calculate the similarity between two documents into the similarity between two vectors. The facial expression features of learners in the situation collected and extracted by the students of this project group are analyzed and modeled, and the time consumption, occupied space, and classification effect of the feature vectors produced by the improved model are integrated. The original feature dimension can be optimized from 100 dimensions to 60 dimensions, which not only saves the time of training feature vectors but also reduces the size of the final feature vectors. Besides, on the basis of 60-dimensional preliminary features extracted by SVM model, four classification models can also achieve the best results. Therefore, in the optimization part of feature extraction, the dimension of initial features extracted by SVM model is set to 60 dimensions. We can gradually use the improved SVM model to analyze the emotional influencing factors and optimization strategies in online teaching, so as to keep abreast of students’ lectures and let more students participate in online teaching as much as possible.

show abstract

Section: Improved Svm Model and Algorithmmentioning

confidence: 99%

Analysis of Affective Factors and Optimization Strategies of Emotion in Online Teaching Based on Improved SVM Model

Huwan²

2022

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…The concurrent speech emotion and naturalness recognition could be approached by utilizing multitask learning: one is to predict valence, arousal, and dominance of the speech signal, and the other is to predict the naturalness score of the same utterance. Indeed, predicting valence, arousal, and dominance simultaneously also can be regarded as a multitask learning problem [8].…”

Section: Introductionmentioning

confidence: 99%

“…Summary of benchmarking results (CCC scores) in this study (MTL and STL) against previous studies (STL). Ref [8]. used subsets of MSP-IMPROV with improvised and natural interaction parts only (MPSIN) using both acoustic and linguistic features.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Speech Emotion and Naturalness Recognitions With Multitask and Single-Task Learnings

Atmaja

Akagi

2022

IEEE Access

Self Cite

View full text Add to dashboard Cite

This paper evaluates speech emotion and naturalness recognitions by utilizing deep learning models with multitask learning and single-task learning approaches. The emotion model accommodates valence, arousal, and dominance attributes known as dimensional emotion. The naturalness ratings are labeled on a five-point scale as dimensional emotion. Multitask learning predicts both dimensional emotion (as the main task) and naturalness scores (as an auxiliary task) simultaneously. The single-task learning predicts either dimensional emotion (valence, arousal, and dominance) or naturalness score independently. The results with multitask learning show improvement from previous studies on single-task learning for both dimensional emotion recognition and naturalness predictions. Within this study, single-task learning still shows superiority over multitask learning for naturalness recognition. The scatter plots of emotion and naturalness prediction scores against the true labels in multitask learning exhibit the lack of the model; it fails to predict the low and extremely high scores. The low score of naturalness prediction in this study is possibly due to a low number of samples of unnatural speech samples since the MSP-IMPROV dataset promotes the naturalness of speech. The finding that is jointly predicting naturalness with emotion helps improve the performance of emotion recognition may be embodied in the emotion recognition model in future work.

show abstract

“…Text-based emotion extraction system tries to extract emotions, such as happiness, fear, and grief, from written text by human [6]. Generally, emotions are complex subjective concepts and fuzzy, that they can express or understand with incorrect form easily [7]. Emotions are too complex in textual dialogues and one of the main reasons in this case relates to facial expressions and the writer sound frequency unavailability [8].…”

Section: Introductionmentioning

confidence: 99%

Emotion Extraction from Text Using Fuzzy-Deep Neural Network

Zaremarjal

Yiltas-Kaplan

Lazemi

2021

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Emotions are one of the most important and affective start points of our decisions in daily life. People express their emotions, consciously or unconsciously, in their behaviors, like speech, facial expressions, textual content, and so on, whether it be in blogs, reviews, or on social media. Due to the rapid growth in textual social media to link people, the emotion extraction from textual data has attracted a lot of attention. This paper performs the sentence-level discrete emotion extraction from text by using Fuzzy Neural Networks (FNN) and Deep Recurrent Neural Networks (DRNN). Much work has been presented to extracting emotion from textual data with precise and constant boundaries, but little of it has considered the inherent uncertainty in natural language. In other words, the high ambiguity of emotion in text data makes an emotionally sentence express multiple emotions at the same time. To deal with this uncertainty we use a Neuro Neural Network (NNN). But from another point of view, by increasing the number of inputs, the number of parameters in NNN growth exponentially. To this end, we extract a low-dimensional semantic representation of input sentence by using a Bi-directional Long Short-Term Memory (Bi-LSTM). Our goal is to achieve better performance with combining the advantages of deep learning in high-level semantic feature extraction to ambiguity handling and the capabilities of fuzzy logic with fuzzy membership degrees to uncertainty handling. Experiments are conducted on the SemEval2007-Task14 dataset that contains 1250 annotated texts with Ekman's basic emotions. The obtained results indicate that our proposed method outperforms the previous methods.

show abstract

Two-stage dimensional emotion recognition by fusing predictions of acoustic and text networks using SVM

Cited by 37 publications

References 37 publications

Analysis of Affective Factors and Optimization Strategies of Emotion in Online Teaching Based on Improved SVM Model

Analysis of Affective Factors and Optimization Strategies of Emotion in Online Teaching Based on Improved SVM Model

Speech Emotion and Naturalness Recognitions With Multitask and Single-Task Learnings

Emotion Extraction from Text Using Fuzzy-Deep Neural Network

Contact Info

Product

Resources

About