学生のNGUYENさんがNCSP'20においてStudent Best paper awardを受賞
学生のNGUYEN, Huy Quocさん（博士前期課程1年、ヒューマンライフデザイン領域・鵜木研究室）がRISP International Workshop on Nonlinear Circuits, Communications and Signal Processing 2020（NCSP'20）において、Student Best paper awardを受賞しました。
Improvement of bone-conducted speech restoration using linear prediction and long short-term memory model
NGUYEN, Huy Quoc
As a solution for speech communication in an extremely noisy environment, bone-conducted (BC) speech has a significant advantage because of its stability against surrounding noise. However, the quality and intelligibility of BC speech are degraded, so the problem of restoring BC speech arises. To solve this problem, we propose a method for restoring BC speech combined of linear prediction (LP) model presented by line spectral frequencies (LSF) and long short-term memory (LSTM) model. We evaluate the proposed method by three objective measurements: log-spectrum distortion, linear prediction coefficient distance, and perceptual evaluation of speech quality. The evaluation results show that our proposed method is better than the previous method that used simple recurrent network in all three measurements. These results also show that the model can yield speech with better quality when the LP gain is estimated more accurately.
This is a great honour for me to receive the Student Paper Award in NCSP'20. For this achievement, I would like to express my gratitude to my supervisor, Prof. Masashi Unoki, and to all colleagues who have supported me with their invaluable comments and guidance without which I would not be able to achieve such award. Finally, I would like to thank my family and friends for supporting and encouraging me all the time.