Design of a Multilingual Translation System with Personalized Voice Characteristics

This study combines AI voice synthesis with IoT to create a real-time multilingual translation system. Front-end devices capture the speaker's voice, which is then played back in both original and translated forms, enabling consecutive interpreting. This system is ideal for multilingual scenarios. After digitization, audio signals undergo speech-to-text conversion, multilingual translation, and text-to-speech processing. Speech synthesis matches the speaker's voice characteristics, allowing listeners to hear the translation in the speaker's voice. Both voice-to-text conversion and translated audio playback use Wi-Fi for data transmission, facilitating future applications. In validation, the average translation and transmission time for each speech segment were 410 ms and 3.38 s, respectively. The synthesized voice similarity analysis achieved an accuracy of 88%, showcasing the method's potential for system integration and development.