Get started
Work

Speech Processing System with Real-Time Emotion and Accent Detection

Year: 2026
Type: preprint
Abstract: Speech processing has become an important com-ponent of modern human–computer interaction systems, en-abling computers to understand, analyze, and respond to spokencommunication. Conventional speech applications generally con-centrate on individual tasks such as speech recognition, voicesynthesis, or classification. However, practical speech-analysisapplications require multiple capabilities to operate together.This paper presents an AI-based speech processing systemthat integrates automatic speech recognition, accent detection,emotion detection, speech analytics, and real-time voice interac-tion within a unified web platform. The proposed system acceptsuploaded or live-recorded speech, performs audio preprocessingand normalization, converts speech into text, detects accent andemotional characteristics, and generates analytical feedback.The frontend is implemented using React.js, while FastAPI isused for backend communication, authentication, asynchronousprocessing, and persistent data management. Machine-learningmodels are incorporated for speech recognition, accent classifi-cation, and emotion analysis. The system additionally calculatesspeech duration, speaking rate, frequently occurring words, andclassification confidence.The proposed architecture provides a unified environment forspeech analysis instead of requiring users to access multipleindependent applications. The system can be applied to languagelearning, communication training, accessibility systems, conver-sational assistants, and speech analytics. This paper presents theliterature survey, research gap, proposed architecture, processingpipeline, machine-learning models, implementation, evaluationmethodology, limitations, and future scope of the system
DOI: 10.5281/zenodo.24140193
Language: English
Open Access status: green
OpenAlex ID: W7225795229
Cites:
Cited by:
Related to: