Automated handcrafted features with deep learning based age group estimation model using facial profiles

• Published in: Multimedia tools and applications 2024-04, Vol.83 (14), p.42149-42164
• Main Authors: Nagaraju, Katta, Reddy, M. Babu
• Format: Article
• Language: English
• Subjects: Age groups; Automation; Belief networks; Classification; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Deep learning; Discriminant analysis; Image filters; Image quality; Machine learning; Multimedia Information Systems; Special Purpose and Application-Based Systems
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-17332-3

At present times, there have been many studies on the automated extraction of facial information using machine learning. Age Group Estimation (AGE) from frontal face images becomes useful in several application areas. The AGE technique aims to classify the age group of the person using the facial image. The stochastic behavior of aging between individuals makes AGE depending upon facial images a tedious process. Faces from distinct age groups have alike features making the facial AGE more difficult. Therefore, this paper considers AGE as a multi-class classification issue and designs an Automated Deep Learning-based Age Group Estimation Model (ADL-AGEM) using facial images. Primarily, the Bilateral Filtering (BF) technique is employed as an image pre-processing technique to boost the facial image quality. In addition, Linear Discriminant Analysis (LDA) technique is applied for the facial component detection process. Besides, a fusion of handcrafted features with deep features takes place to derive a useful set of feature vectors from the facial images. The Local Diagonal Extreme Pattern (LDEP) based handcrafted and Inception v3-based deep features are fused for a facial component before the classification process. In the final stage, the Deep Belief Network (DBN) model is applied as the classifier to determine the appropriate age groups for the applied facial images. To validate the effectiveness of the ADL-AGEM model, a set of experimentations take place on three benchmark databases. The experimental results ensured that the ADL-AGEM model has accomplished promising results over the existing techniques in terms of different measures.