We are conducting research on a lung sound test system that can easily detect lung sounds at home and detect abnormalities at an early stage, such as a body fat meter or blood pressure monitor. We research a method to detect abnormal sounds from characteristics differences between abnormal sounds (called side noise) and normal lung sounds (using fluctuations in the frequency and energy of lung sounds).

We research methods to estimate emotions such as “anger” and “sadness” from differences in the characteristics of baby crying. We use information on how childcare experts hear crying and make decisions because it is difficult to know the true emotions of a baby. Being able to know a little about the emotions when a baby is crying can help new moms and dads.

Free speech is speech that we speak with others emotionally in our daily lives. Due to the development of speech recognition technology, the accuracy of speech recognition has greatly improved, and it has become possible to recognize speech utterances of TV newscasters and voice commands for operating machines with extremely high accuracy. On the other hand, there are many issues to be solved for high-precision recognition in free speech. We research a recognition system that takes into account the unique grammar, words, speakers, and utterances in order to achieve high-accuracy speech recognition. We aim to automatically create meeting minutes and things like that.

• Encouraging Prize (IEICE, Kyusyu Section)
  Keishi Fukunaga：

  Construction of speaker cluster for recognition considering speaker characteristics of conference speech

  Naoki Umeno：

  Extension of Learning Data Considering Similarity of Auscultation Sound Spectrum for Detection of Abnormal Lung Sound

  Wint Yee：

  Acoustic Features for Detection of Emotion in Infant Cry and Abnormal Lung Sound
• Best Paper Award (The 10th International Conference on Science and Engineering 2019)
  Wint Yee：

  Effective Acoustic Feature Parameters for Emotion Estimation of Infant Crying Sound