NAO vs. Pepper: Speech Recognition Performance Assessment

Social robots are becoming increasingly popular due to their communication capabilities in various fields, such as schools, hospitals and other service industries. However, sometimes, it can be challenging to understand a person’s voice due to background noise or auditory problems. To ensure effective communication, robots must have a good understanding of human speech. In order to evaluate their speech comprehension, an exploration of a speech recognition system is required. The present study focuses on the speech recognition system of two social robots, NAO and Pepper. The study aims to compare the robots’ speech recognition systems and determine which performs the best. A speech-to-text conversion tool, Whisper, is integrated with robots’ speech recognition systems to achieve this goal. Furthermore, evaluation measures such as WER, MER, WIL and CER are employed to determine the discrepancies between the original and the recorded speech by analyzing the corresponding text. The findings of the present paper suggest that both NAO and Pepper robot’s speech recognition systems performed equally well, however, additional screen to display spoken words makes Pepper more valuable.