Improve the Traditional Autoencoder Student Classroom Behavior Recognition Algorithm

Most of the existing behavior recognition methods are aimed at the dynamic behavior in action. When applied to the static behavior recognition environment in the classroom, it is difficult to achieve the ideal effect and waste computing resources. Therefore, aiming at the specific environment of cla...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2022 2nd International Conference on Computational Modeling, Simulation and Data Analysis (CMSDA) S. 234 - 239
Hauptverfasser: Zhang, Man, Wei, Yan
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 02.12.2022
Schlagworte:
Analytical models
Autoencoder
Behavioral sciences
Classroom behavior recognition
Computational modeling
Data models
Deep learning
Feature extraction
Loading
Training
VAE
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Most of the existing behavior recognition methods are aimed at the dynamic behavior in action. When applied to the static behavior recognition environment in the classroom, it is difficult to achieve the ideal effect and waste computing resources. Therefore, aiming at the specific environment of classroom, the paper firstly describes the model theories of traditional Auto Encoder (AE) and Variational Auto Encoder (VAE). On this basis, the optimized variational autoencoder (AVAE) model structure proposed in this paper for the classroom behavior recognition of students is introduced. By loading the encoder with pre-training weight, the feature extraction results are obtained, and the loss function is optimized to reduce the supervision of the decoder during training, so as to identify the student behavior more efficiently. The AVAE model is compared with the current common deep learning feature extraction classification methods, and the performance advantages of the proposed method are verified by experiments.
DOI:10.1109/CMSDA58069.2022.00049