Implementation of MFCC Features Extraction with Evaluation of Recurrent Neural Networks and Bidirectional LSTM for Speech to Text Transcription: A Case Study on the Lampung Language Dialect Api

DOI

10.2991/978-94-6463-730-4_13How to use a DOI?

Keywords

MFCC; RNN; BiLSTM; Speech to Text; Lampung Language

Abstract

Lampung language is a cultural heritage of Lampung province which is currently sought to be preserved and maintained its existence amid the declining trend in the use of Lampung language in the Lampung community itself. This research aims to encourage the need for preservation and widespread and inclusive use of Lampung language through speech-to-text conversion technology by utilizing MFCC feature extraction as well as RNN and BiLSTM models and evaluating the performance of acoustic models in text transcription from voice signals, using Word Error Rate (WER) as the main metric. The dataset consisted of regional language-based transcriptions, which were processed through the RNN and BiLSTM models to obtain text output from Lampung language voice conversion using dialect Api. This study used 800 audio sentences in wav format from 4 respondents of Lampung ethnic origin. During the prediction process, the accuracy of the RNN model produced a WER value of 0.5857 and BiLSTM showed a WER of 0.5589. The comparison of these two types of machine learning models shows that BiLSTM produces slightly better model results. Furthermore, the acoustic model was also implemented with 98.53% accuracy on the validation data after 20 epochs, indicating the model learned the data well. Despite the high accuracy, the main challenge faced in these models is generating optimal readable transcriptions, which is due to decoding and tokenization issues.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - M. Ramadhan
AU  - Akmal Junaidi
AU  - Aristoteles
AU  - Favorisen R. Lumbanraja
AU  - Ahmad Faisol
PY  - 2025
DA  - 2025/05/27
TI  - Implementation of MFCC Features Extraction with Evaluation of Recurrent Neural Networks and Bidirectional LSTM for Speech to Text Transcription: A Case Study on the Lampung Language Dialect Api
BT  - Proceedings of the 5th International Conference on Applied Sciences, Mathematics, and Informatics (ICASMI 2024)
PB  - Atlantis Press
SP  - 139
EP  - 150
SN  - 2352-541X
UR  - https://doi.org/10.2991/978-94-6463-730-4_13
DO  - 10.2991/978-94-6463-730-4_13
ID  - Ramadhan2025
ER  -