Enhancing GAIT Analysis: Integrating Trickster Coyote Optimization With FL-BiLSTM Classifier For Exclusion Of Carried Objects

In this research paper, an innovative approach for emotion classification is presented using a hybrid model called Augmented Trickster Coyote-based Federated Learning Bidirectional Long Short-Term Memory (FL-BiLSTM) classifier. The hyper parameters of the FL-BiLSTM classifier are optimized using the Trickster Coyote Optimization (TCO) algorithm, resulting in improved classification accuracy. The effectiveness is demonstrated by extensive experiments and comparisons with existing methods. Introduction: Emotion recognition plays a crucial role in various applications, including human-computer interaction and affective computing. Federated Learning (FL) provides a privacy-friendly approach for training deep learning models on distributed datasets. However, the performance of FL models can be affected by a suboptimal choice of hyper parameters. Objectives: - Create a unique way to emotion recognition utilizing a hybrid model, and improve the performance of Federated Learning-based emotion categorization by optimizing the hyper parameters. Extensive trials and comparisons with existing approaches will be used to demonstrate the optimized model's efficiency. Methods:- We suggest the Federated Learning Bidirectional Long Short-Term Memory (FL-BiLSTM) classifier for emotion classification. Apply the Trickster Coyote Optimization (TCO) method to optimize the FL-BiLSTM classifier's hyper parameters. Employ Federated Learning to train an emotion classification model on distributed data while maintaining anonymity.- Conduct studies that contrast the efficacy of the suggested model with other methods that are currently in use. Results: The suggested Augmented Trickster Coyote-based FL-BiLSTM model outperforms previous techniques in terms of classification accuracy.The TCO method successfully improves the FL-BiLSTM's hyperparameters that resulting in higher scores on recognizing emotions tests.The sensitivity of the trickster coyote-based FL-BiLSTM obtains a noteworthy 98.800% when the training rate is set to 80% after 100 epochs. The trickster coyote-based FL-BiLSTM has a specificity of 89.198% & a training percentage of 70% after 100 epochs. Conclusions: The hybrid model, which combines Federated Learning, BiLSTM, and TCO, provides a promising method to identify emotions with higher accuracy & privacy-preserving features. The TCO method optimizes hyperparameters, improving the performance of the FL-BiLSTM classifier, making it a useful tool for distributed emotion classification applications. The proposed approach surpasses existing models, indicating that it has potential applications in human-computer interaction or social computer systems.