EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Speech Analysis Language Identification and Translation

G. Ramya, N. Chandra Lekha, P. Pranathi, P. Sahana

International Journal of Advances in Artificial Intelligence and Machine Learning · 2025

DOI: 10.58723/ijaaiml.v2i3.461 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Advances in Artificial Intelligence and Machine Learning, 2025. Verfasst von G. Ramya und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background of study: The increasing globalization of communication has intensified the need for systems capable of automatically identifying spoken languages and providing accurate, real-time translation. With advancements in speech processing and machine learning, an integrated framework for speech analysis, language identification, and translation has become both feasible and necessary.

Aims

This paper aims to develop and evaluate a comprehensive system that performs automatic speech preprocessing, language identification, speech recognition, and machine translation. The study focuses on designing a multilingual pipeline capable of detecting multiple languages, converting speech to text, and translating the output into a target language with high accuracy and usability.

Methods

A multilingual speech corpus comprising recordings in English, Spanish, French, and Mandarin was used. Audio underwent preprocessing, feature extraction using MFCCs and spectrograms, and language identification using CNN-based MFCC classifiers as well as i-vector and x-vector models. Speech recognition was conducted using pre-trained ASR systems such as Whisper and DeepSpeech, followed by neural machine translation (NMT). System performance was evaluated through accuracy, precision, recall, BLEU scores, real-time factor (RTF), and user experience assessments.

Result

The proposed system demonstrated strong performance across the LID, ASR, and translation components. CNN-based language identification achieved high accuracy across multilingual inputs, while ASR models produced coherent transcriptions suitable for downstream translation. Translation evaluation using BLEU scores and qualitative human review confirmed that the pipeline maintained contextual accuracy. The system also showed robustness across varying speakers, accents, and noise conditions.

Conclusion

The integrated Speech Analysis, Language Identification, and Translation system provides an effective solution for overcoming language barriers in real-time communication. By combining noise-reduced audio preprocessing, reliable language detection, and accurate translation, the system offers a user-friendly platform suitable for multilingual applications. Future improvements include expanding the language set, enhancing robustness against dialectal variation, and deploying the model on lightweight edge devices for real-time applications.

Zitieren

APA

Ramya, G., Lekha, N. C., Pranathi, P., & Sahana, P. (2025). Speech Analysis Language Identification and Translation. International Journal of Advances in Artificial Intelligence and Machine Learning. https://doi.org/10.58723/ijaaiml.v2i3.461

Vancouver

Ramya G, Lekha NC, Pranathi P, Sahana P. Speech Analysis Language Identification and Translation. International Journal of Advances in Artificial Intelligence and Machine Learning. 2025. doi:10.58723/ijaaiml.v2i3.461

BibTeX
@article{Ramya2025,
  author = {Ramya, G. and Lekha, N. Chandra and Pranathi, P. and Sahana, P.},
  title = {Speech Analysis Language Identification and Translation},
  journal = {International Journal of Advances in Artificial Intelligence and Machine Learning},
  year = {2025},
  doi = {10.58723/ijaaiml.v2i3.461}
}
Kostenlos starten →