Research Topics

Japanese | English ]

Speech and Language

As speech is the most natural way for humans to communicate, spoken language interfaces are considered to be the best modality for a wide range of information systems. We are investigating acoustic modeling, noise suppression systems to improve speech recognition performance, and communication analysis for application to dialog system.


We often perform tasks such as identifying the location of a detected sound or grasping the size of a room using both ears. 
These abilities, studied in the field of Acoustics, are the basis for our work, which has potential application in, for example, virtual reality systems and computer games. The main tool for the analysis of sound localization is the head-related transfer function (HRTF). HRTF contains reflections and refraction of sound waves at the ears and head. By using the HRTF we can, for example, place the sound source at arbitrary locations.

<<for more detail


Our group focuses on the relationship between lingual characteristic of lyrics or nationality of composer, music composition, and score following system. For such researches, we have been using signal processing and text analysis techniques. Since multimedia information is combined, song retrieval can also be done based on both keywords and situational factors.

<<for more detail

Driving Behavior

The driving behavior signal processing group focuses on stochastic modeling of the driver, the vehicle, and the environment as a single system. We apply signal processing techniques to multimodal driving data recorded by driving simulators and a real instrumented vehicle.These are designed and built in our lab. And we also investigate above driving data compression and driving scene recognition system.

>> for more detail