Utilizing Temporal Inductive Path Neural Networks for Accurate Voice-Based Depression Classification: A Detailed...

Depressive disorder is a common mental illness often triggered by environmental stress or social influences. It impacts mood and behavior, leading to difficulties in areas such as education, family life, and the workplace. In severe cases, depression can result in suicide attempts. Fortunately, depression is a treatable condition when properly diagnosed by mental health professionals. However, many individuals aware of their mental health issues hesitate to seek psychiatric care due to long waiting times and high treatment costs. To overcome this problem, this paper proposes a voice-based depression classification using temporal inductive path neural networks (TIPNNs) for analyzing speech patterns to identify mental health states (AVA-TIPNN-DD). Initially, input data are gathered from voice recordings. The input data are preprocessed using the Koopman Kalman particle filter, which is used to clean the audio recordings. After that, the preprocessed data are given to multi objective matched synchro squeezing chirplet transform for extracting the spectrum features such as power spectrum density, spectral power, spectral centroid, and spectral flatness measure. Then the extracted features are fed into TIPNN, which is used for detecting and classifying depressions such as depressed and nondepressed. Generally, TIPNN does not show adapting optimization strategies to determine optimal parameters to ensure accurate detection. Hence, the binary battle royale optimizer algorithm is used to optimize the weight parameter of TIPNN, which accurately detects depression. The proposed method is implemented in Python, and the efficacy of the AVA-TIPNN-DD technique was assessed based on various performance measures, including Accuracy, Precision, Sensitivity, Specificity, and F1-score. Performance of the AVA-TIPNN-DD method achieves 99.26% Accuracy, 99.5% Precision, 99.6% Sensitivity, and 99.5% F1-Score when analyzed through existing techniques such as the decision-support system for major depression detection utilizing spectrum and convolution neural network with electroencephalogram signals (DSS-DD-CNN), a textual-based feature method for depression recognition by machine learning classifiers and social media texts (TBFA-DD-FFANN), and smart voice detection that relies on deep learning for depression diagnosis (SVR-DD-DRN), respectively.