出身大学 【 表示 ／ 非表示 】

出身大学 【 表示 ／ 非表示 】

出身大学院 【 表示 ／ 非表示 】

出身大学院 【 表示 ／ 非表示 】

取得学位 【 表示 ／ 非表示 】

取得学位 【 表示 ／ 非表示 】

学内職務経歴 【 表示 ／ 非表示 】

学内職務経歴 【 表示 ／ 非表示 】

学外略歴 【 表示 ／ 非表示 】

学外略歴 【 表示 ／ 非表示 】

全件表示 >>

所属学会・委員会 【 表示 ／ 非表示 】

所属学会・委員会 【 表示 ／ 非表示 】

専門分野（科研費分類） 【 表示 ／ 非表示 】

専門分野（科研費分類） 【 表示 ／ 非表示 】

研究経歴 【 表示 ／ 非表示 】

研究経歴 【 表示 ／ 非表示 】

• アナログ集積回路

  アナログ集積回路  

  研究期間： 2010年04月  -  継続中

• 再構成可能なデータコンバータ

  再構成可能なデータコンバータ回路  

  研究期間： 2008年04月  -  継続中

• 生体信号センサーの開発

  味覚センサー,システムLSI,アナログ回路  

  研究期間： 2007年04月  -  継続中

• 確率共鳴法による信号処理

  信号処理,確率共鳴,アナログ回路  

  研究期間： 2015年04月  -  継続中

論文 【 表示 ／ 非表示 】

論文 【 表示 ／ 非表示 】

• Discussion on differences in frequency band required for horizontal sound localization by age

  Okuno T., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

  ACM International Conference Proceeding Series      2019年05月  [査読有り]

   概要を見る

  © 2019 Association for Computing Machinery. In this paper, the horizontal sound image experiment of the real sound source was performed under the head resting condition using the band - limited noise using the low pass filter and the band pass filter. Then, the effect of each frequency band on direction perception of the hearing person and the elderly person was clarified. Five people with healthy men in their twenties and five elderly people in their 70s male hearing white noise pass through low-pass filters (2, 3, 4, 5, 6 kHz) and band-pass filters (1-4, 2-4, 2-5 [kHz]) passed through the directional perception experiment. As a result, similarly to the previous study [1], it is difficult to perceive directions sufficiently if two or more frequency bands are included, and it is difficult to perceive the direction of band-pass noise through a band-pass filter of a narrow frequency band. As a result of this research, Direction perception was found that the frequency band around 4 [kHz] and 6 [kHz] is small as an element of direction perception and the frequency band around 5 [kHz] is an important clue for direction perception.

  DOI Scopus

• Enhancement of electrolarynx speech based on WaveRNN

  Urabe E., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

  ACM International Conference Proceeding Series      2019年05月  [査読有り]

   概要を見る

  © 2019 Association for Computing Machinery.ACM In Japan, there are about 30,000 people whose larynxes have been removed. These people are not able to make voice, so they have problems in communication. For example, they cannot communicate their will enough with communication by writing or gestures alone, thereby avoiding communication. However, they can move the mouth and tongue normally. Therefore, they speak using alternative vocalizations. Electrical artificial larynx (EL) is one of the alternative vocalizations. When the vibrator is pressed against the lower part of the jaw, the vibration sound is transmitted in the sound path. This speech method is easy to learn and is cleaner than other speech methods, but this voice sound is bad quality because it has no intonation and it is noisy. In this study, we tried to improve the EL sound quality by using WaveRNN that is one of Deep learning. After making WaveRNN learn human voice, we converted the EL voice. We evaluated the converted voice in a subjective evaluation experiment in 10 words that selected each consonant and vowels in a balanced manner. The evaluation items are: 1. Clarity of consonants, 2. Clarity of vowels, 3. Clarity of whole voice, 4. Natural, 5. Easy to hear. As a result, Clarity and naturalness of consonants improved in words containing two or more unvoiced consonants. On the other hand, in the voiced consonants and vowels, no improvement was seen by the proposed method. Therefore, we think that voice quality was improved in unvoiced consonants. On the other hand, there was no tendency for improvement in voiced consonants and vowels.

  DOI Scopus

• Examination of acoustic features for discriminating between real and loudspeaker speeches

  Kinoshita Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

  ACM International Conference Proceeding Series      2019年05月  [査読有り]

   概要を見る

  © 2019 Association for Computing Machinery. In the train while driving, there are many noises such as noises from the vibration of the vehicle and the shock noise generated at the vehicle connection, and the roar of human beings, etc. In addition to stressing the passengers, these noises it prevents you from listening to the announcements in the train. In this paper, a new noise canceling earphone system aimed at improving in-car announcement speech listening and reducing stress that the noise gives to passengers by separating surrounding speech and in-train announcement speech regarded as noise I devised. In this study, as a preliminary step to complete the system, in order to find the difference between the ambient speech and the announcement speech such as the roar of the passenger, both speeches were compared using acoustic features. As a result, among the acoustic features used in the comparison, differences were found in the spectral centroid, the spectral roll-off, and the intensity and spectral contrast when divided into sub-bands.

  DOI Scopus

• In load collapse prevention system Examination of the relationship between truck vibration and load inclination

  Inoshita C., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

  ACM International Conference Proceeding Series      2019年05月  [査読有り]

   概要を見る

  © 2019 Association for Computing Machinery. ACM In recent years, the demand for the logistics industry has increased enormously due to the spread of Internet mail order. On the other hand, the number of fatal accidents caused by trucks is more than 10,000 a year in Japan. Above all, there is a problem that there is an annual fatal and fatal accident due to truck load collapse. In this study, using the drive recorder, we calculated the inclination of the package photographed during driving by using KCF of object tracking method of image processing. Comparing the change of the Luggage tilt and the amount of change in 3-axis acceleration attached to the vehicle/ time of each frame in all the data, frames at each maximum value point almost agreed at the time of” When riding over convex part of inclined road”. On the other hand, “When a sudden steering wheel during driving”, a point where the absolute value of the change in package inclination during one frame reaches 0.005 ° and the point where the maximum change occurs after that. It almost coincided with the point where the curve was reached in the video. As for the 3-axis acceleration sensor, there is no difference in tendency between the curve and the straight running, and it is considered as the main cause that the influence of the roughness of the road is greatly influenced rather than the influence at the time of the curve.

  DOI Scopus

• Study on evaluating risks of routes damages at Earthquakes for evacuation guidance system

  Mihara Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

  ACM International Conference Proceeding Series      2019年05月  [査読有り]

   概要を見る

  © 2019 Association for Computing Machinery. In recent years, the damage caused by the earthquake has become very large in Japan. Secondary disasters such as tsunami and fire are the main causes of death due to the earthquake, both of which are causes of evacuation delay. Rapid evacuation is important in the event of a disaster. However, due to the earthquake, various obstacles will occur in the aisle. Obstacles may occur due to the collapse of houses due to the earthquake and it may take time to pass. If you can obtain information about the aisle in advance and avoid evacuation routes in advance, you can complete evacuation safer and more quickly. As a method of acquiring traffic information, We thought that the use of SNS is useful from the viewpoint of the real-time nature of information and the penetration rate of young people aggressively actively communicating information. In this research, we proposed an evacuation guidance system to evacuate users with a safe shortest route. Then we examined the danger of passing through obstacles. As a result, "road shredder" is evaluated as the most dangerous, and "road crack" is evaluated as the most secure.

  DOI Scopus

全件表示 >>

口頭発表・ポスター発表等 【 表示 ／ 非表示 】

口頭発表・ポスター発表等 【 表示 ／ 非表示 】

• Study on Mistype Correction Support Using Attention in Japanese Input

  Ryuki Komatsu, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

  2020 IEEE 38th International Conference on Consumer Electronics  (Las Vegas, U.S.A.)  2020年01月  -  2020年01月   

• Electrolarynx System Using Voice Conversion Based on WaveRNN

  Etsuro Urabe, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

  2020 IEEE 38th International Conference on Consumer Electronics  (Las Vegas, U.S.A.)  2020年01月  -  2020年01月   

• 電車内アナウンス音声と周囲の会話を判別するための音響的特徴量の検討

  木下雄斗,中藤良久,平川凛,中司賢一,河野英昭

  第72回電気・情報関係学会九州支部連合大会  (北九州市)  2019年09月  -  2019年09月   

• インバータによる低消費電力高利得アンプの開発

  塩見涼陽,中司賢一

  第72回電気・情報関係学会九州支部連合大会  (北九州市)  2019年09月  -  2019年09月   

• アルコールの近赤外吸収特性の2波長での測定

  永吉賢,中司賢一

  第72回電気・情報関係学会九州支部連合大会  (北九州市)  2019年09月  -  2019年09月   

全件表示 >>

科研費獲得実績 【 表示 ／ 非表示 】

科研費獲得実績 【 表示 ／ 非表示 】

• 感性バイオセンサの開発

  基盤研究(S)

  研究期間:  2006年04月  -  2011年03月

  研究課題番号:  18106009

海外研究歴 【 表示 ／ 非表示 】

海外研究歴 【 表示 ／ 非表示 】

• Low-power, high-efficiency power supply systems and low-power integrated circuits design

  University of California at Berkeley  アメリカ合衆国  研究期間:  2000年04月  -  2001年03月

担当授業科目 【 表示 ／ 非表示 】

担当授業科目 【 表示 ／ 非表示 】

全件表示 >>