Article: Development of hybrid convolution (1D/2D)-based Conv-LSTM with attention mechanism for predicting health disorders associated with obesity using photoplethysmography signal Journal: International Journal of Bioinformatics Research and Applications (IJBRA) 2026 Vol.22 No.3 pp.233 - 265 Abstract: A deep learning-based obesity prediction model is developed using photoplethysmography (PPG) signal features. The required PPG signals are gathered from the available data resource. The gathered PPG signals are fed in to signal preprocessing. Here the motion aircrafts (MA) removal and filtering approach are utilised and further it is split into distinct windows. Based on the signal split window, the spectral feature are extracted which is denoted as F1. Similarly, the PPG signal is given to the adaptive short-time Fourier transform (ASTFT), which generates the spectrogram image that is designated as F2. Some of the parameters in the STFT are tuned by using the enhanced energy valley optimiser (EEVO). Finally, both the features are subjected to hybrid convolution (1D-2D)-based convolutional long short-term memory with Attention mechanism (HC-Conv-LSTM-AM). The functionality of the proposed obesity prediction model is compared with existing approaches. The output defined the developed approach outperformed the traditional models. Inderscience Publishers - linking academia, business and industry through research

Title: Development of hybrid convolution (1D/2D)-based Conv-LSTM with attention mechanism for predicting health disorders associated with obesity using photoplethysmography signal

Authors: Venumaheswar Rao Bondala; Ashoka Reddy Komalla

Addresses: Department of E&I Engineering, Kakatiya Institute of Technology and Science, Kakatiya University Warangal, Telangana, India ' Department of ECE, Kakatiya Institute of Technology and Science, Warangal, Telangana, India

Abstract: A deep learning-based obesity prediction model is developed using photoplethysmography (PPG) signal features. The required PPG signals are gathered from the available data resource. The gathered PPG signals are fed in to signal preprocessing. Here the motion aircrafts (MA) removal and filtering approach are utilised and further it is split into distinct windows. Based on the signal split window, the spectral feature are extracted which is denoted as F1. Similarly, the PPG signal is given to the adaptive short-time Fourier transform (ASTFT), which generates the spectrogram image that is designated as F2. Some of the parameters in the STFT are tuned by using the enhanced energy valley optimiser (EEVO). Finally, both the features are subjected to hybrid convolution (1D-2D)-based convolutional long short-term memory with Attention mechanism (HC-Conv-LSTM-AM). The functionality of the proposed obesity prediction model is compared with existing approaches. The output defined the developed approach outperformed the traditional models.

Keywords: health disorders; obesity prediction; hybrid convolution-based convolutional long short-term memory with attention mechanism; HC-Conv-LSTM-AM; enhanced energy valley optimiser; EEVO; adaptive short-time Fourier.

DOI: 10.1504/IJBRA.2026.154824

International Journal of Bioinformatics Research and Applications, 2026 Vol.22 No.3, pp.233 - 265

Received: 21 May 2024
Accepted: 24 Mar 2025
Published online: 15 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Development of hybrid convolution (1D/2D)-based Conv-LSTM with attention mechanism for predicting health disorders associated with obesity using photoplethysmography signal

Keep up-to-date