Wind turbine blades can be damaged by lightning strikes, flying debris, and long-term deterioration. Since the tip speed of the blades exceeds 200 km/h, severe accidents may occur if the damage progresses and causes blade failure. Currently, blade inspections are mainly conducted through visual inspections after stopping the turbine rotation. However, inspections are infrequent, and small damages are difficult to detect from ground-based visual observations. Therefore, we are conducting research on  damage detection technologies using sound radiated from rotating blades. This approach aims to realize continuous monitoring without stopping the turbines and to enable early detection of small damages. Furthermore, our research focuses not only on detecting the presence of damage, but also on estimating the damage location and severity.

During disasters, rapid search and rescue operations are required. Drones have attracted attention because they can be deployed immediately and can efficiently perform search operations from the air. Currently, many drone-based search methods rely on cameras. However, it is difficult to detect victims at night under low-light conditions or victims trapped under rubble. Therefore, we are conducting research on technologies for estimating the locations of victims using sounds such as human voices captured by multiple microphones mounted on drones. In this research, acoustic signal processing under the strong ego-noise generated by drones becomes a major challenge.

During disaster search and rescue operations, rapid understanding of ground conditions is required in addition to victim search in order to assess disaster situations and secure rescue routes. Currently, cameras and LiDAR are widely used for ground surface sensing. However, when considering simultaneous operation with acoustic-based victim search, there are limitations in the weight of onboard equipment and computational cost for drones. Therefore, we are conducting research on technologies for sensing ground surface shapes and road damage using reflected sounds from the ground with hardware and signal processing shared with acoustic-based victim search. This technology is expected to be applicable not only to disaster response, but also to infrastructure inspection and robot localization and mapping.

With the progression of an aging society, accidents caused by elderly people falling at home are increasing. In particular, for elderly people living alone, the automation of fall detection is becoming increasingly important due to the shortage of caregivers. Cameras can be considered as a method for monitoring living spaces. However, their use is difficult in spaces where privacy considerations are required, such as toilets and bathrooms. Therefore, we are conducting research on technologies for estimating human position and posture using sound while preserving privacy. Reverberation is generally a problem for acoustic measurement in narrow spaces such as toilets and bathrooms. In this research, reverberation is instead utilized to estimate human position and posture from reverberation distributions for fall detection.