Kalmora: A Voice-Based Journaling App for Real-Time Emotion Detection and Sustainable Mental Well-Being

Abstract

The current tools for journaling depend on personal self-reporting which fails to match accurately with how people genuinely feel, and emotional states affect sustainable societal development. This research introduces Kalmora, which stands as a mobile voice-journaling application which utilizes Wav2Vec2 speech emotion recognition model that identifies seven basic emotions (happiness, sadness, anger, fear, disgust, neutral, surprise) in real time. Kalmora's secure dual frontend backend framework consisting of Flutter and Flask and Firebase elements performs time-based emotion assessment and individual wellness guidance. The model evaluated using controlled TESS data reached 99.8% accuracy which surpassed CNN-LSTM benchmark models at 94.1% accuracy. User testing involved observing real users interacting with the app to evaluate the ease of voice journaling, accuracy of emotion detection, and overall user experience, leading to improvements based on their feedback. Through its combination of objective emotional knowledge and practical tips Kalmora brings new possibilities to digital mental healthcare that enable sustainable emotional self-care practices.