Novel Task-Based Unification and Adaptation (TUA) Transfer Learning Approach for Bilingual Emotional Speech Data

Modern developments in machine learning methodology have produced effective approaches to speech emotion recognition. The field of data mining is widely employed numerous situations where it possible predict future outcomes by using the input sequence from previous training data. Since feature space and distribution are same for both testing conventional approaches, they drawn pool. However, because so many applications require a difference data, gathering becoming more expensive. High performance learners that been trained similar, already-existing needed these situations. To increase model’s capacity learning, transfer involves transferring knowledge one domain another related domain. address this scenario, we extracted ten multi-dimensional features signals OpenSmile method classify various datasets. In paper, emphasize importance novel system called Task-based Unification Adaptation (TUA), which bridges disparity between extensive upstream downstream customization. We take advantage two components TUA, task-challenging unification task-specific adaptation. Our algorithm studied following datasets: Arabic Emirati-accented dataset (ESD), English Speech Under Simulated Actual Stress (SUSAS) Ryerson Audio-Visual Database Emotional Song (RAVDESS). Using multidimensional on given datasets, were able achieve an average recognition rate 91.2% ESD, 84.7% RAVDESS 88.5% SUSAS respectively.