Speech-to-Text Conversion in Indonesian Language Using a Deep Bidirectional Long Short-Term Memory Algorithm

Nowadays, speech is used also for communication between humans and computers, which requires conversion from to text. Nevertheless, few studies have been performed on speech-to-text in Indonesian language, most were limited the of datasets with incomplete sentences. In this study, complete sentences language using deep bidirectional long short-term memory (LSTM) algorithm. Spectrograms Mel frequency cepstral coefficients (MFCCs) utilized as features a total 5000 data spoken by ten subjects (five males five females). The results showed that LSTM algorithm successfully converted text Indonesian. accuracy achieved MFCC was higher than spectrograms; obtained best word error rate value 0.2745% while spectrograms 2.0784%. Thus, MFCCs are more suitable feature study will help implementation tools other languages.