Improving CNN-based solutions for emotion recognition using evolutionary algorithms
AI-based approaches, especially deep learning have made remarkable achievements in Speech Emotion Recognition (SER). Needless to say, Convolutional Neural Networks (CNNs) have been the backbone of many of these solutions. Although the use of CNNs have resulted in high performing models, building the...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-05-01
|
| Series: | Results in Applied Mathematics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590037423000067 |
| _version_ | 1797815485434167296 |
|---|---|
| author | Parsa Mohammadrezaei Mohammad Aminan Mohammad Soltanian Keivan Borna |
| author_facet | Parsa Mohammadrezaei Mohammad Aminan Mohammad Soltanian Keivan Borna |
| author_sort | Parsa Mohammadrezaei |
| collection | DOAJ |
| description | AI-based approaches, especially deep learning have made remarkable achievements in Speech Emotion Recognition (SER). Needless to say, Convolutional Neural Networks (CNNs) have been the backbone of many of these solutions. Although the use of CNNs have resulted in high performing models, building them needs domain knowledge and direct human intervention. The same issue arises while improving a model. To solve this problem, we use techniques that were firstly introduced in Neural Architecture Search (NAS) and use a genetic process to search for models with improved accuracy. More specifically, we insert blocks with dynamic structures in between the layers of an already existing model and then use genetic operations (i.e. selection, mutation, and crossover) to find the best performing structures. To validate our method, we use this algorithm to improve architectures by searching on the Berlin Database of Emotional Speech (EMODB). The experimental results show at least 1.7% performance improvement in terms of Accuracy on EMODB test set. |
| first_indexed | 2024-03-13T08:23:25Z |
| format | Article |
| id | doaj.art-1ba2574996df485fa9e192a8b367a373 |
| institution | Directory Open Access Journal |
| issn | 2590-0374 |
| language | English |
| last_indexed | 2024-03-13T08:23:25Z |
| publishDate | 2023-05-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Results in Applied Mathematics |
| spelling | doaj.art-1ba2574996df485fa9e192a8b367a3732023-05-31T04:47:30ZengElsevierResults in Applied Mathematics2590-03742023-05-0118100360Improving CNN-based solutions for emotion recognition using evolutionary algorithmsParsa Mohammadrezaei0Mohammad Aminan1Mohammad Soltanian2Keivan Borna3Faculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, IranFaculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, IranFaculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, IranCorresponding author.; Faculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, IranAI-based approaches, especially deep learning have made remarkable achievements in Speech Emotion Recognition (SER). Needless to say, Convolutional Neural Networks (CNNs) have been the backbone of many of these solutions. Although the use of CNNs have resulted in high performing models, building them needs domain knowledge and direct human intervention. The same issue arises while improving a model. To solve this problem, we use techniques that were firstly introduced in Neural Architecture Search (NAS) and use a genetic process to search for models with improved accuracy. More specifically, we insert blocks with dynamic structures in between the layers of an already existing model and then use genetic operations (i.e. selection, mutation, and crossover) to find the best performing structures. To validate our method, we use this algorithm to improve architectures by searching on the Berlin Database of Emotional Speech (EMODB). The experimental results show at least 1.7% performance improvement in terms of Accuracy on EMODB test set.http://www.sciencedirect.com/science/article/pii/S2590037423000067Convolutional Neural NetworksDeep learningGenetic algorithmsEvolutionary optimization |
| spellingShingle | Parsa Mohammadrezaei Mohammad Aminan Mohammad Soltanian Keivan Borna Improving CNN-based solutions for emotion recognition using evolutionary algorithms Results in Applied Mathematics Convolutional Neural Networks Deep learning Genetic algorithms Evolutionary optimization |
| title | Improving CNN-based solutions for emotion recognition using evolutionary algorithms |
| title_full | Improving CNN-based solutions for emotion recognition using evolutionary algorithms |
| title_fullStr | Improving CNN-based solutions for emotion recognition using evolutionary algorithms |
| title_full_unstemmed | Improving CNN-based solutions for emotion recognition using evolutionary algorithms |
| title_short | Improving CNN-based solutions for emotion recognition using evolutionary algorithms |
| title_sort | improving cnn based solutions for emotion recognition using evolutionary algorithms |
| topic | Convolutional Neural Networks Deep learning Genetic algorithms Evolutionary optimization |
| url | http://www.sciencedirect.com/science/article/pii/S2590037423000067 |
| work_keys_str_mv | AT parsamohammadrezaei improvingcnnbasedsolutionsforemotionrecognitionusingevolutionaryalgorithms AT mohammadaminan improvingcnnbasedsolutionsforemotionrecognitionusingevolutionaryalgorithms AT mohammadsoltanian improvingcnnbasedsolutionsforemotionrecognitionusingevolutionaryalgorithms AT keivanborna improvingcnnbasedsolutionsforemotionrecognitionusingevolutionaryalgorithms |