1.3 Application of HARK

We have developed an auditory function robot with binaural hearing using two microphones and used three-speaker simultaneous speech recognition as a kind of benchmark. For auditory functions on SIG and SIG2 (upper body humanoid robots), simultaneous speech recognition at 1m away from three speakers who stand at an interval of 30 degrees is possible at a certain level of accuracy [16]. However, this system requires a large amount of prior knowledge and processing and therefore we had to judge that it would be difficult to equip this system as an easy auditory function that can be used in sound environments. In order to overcome this performance limit, we started to develop an auditory function robot with an increased number of microphones and invented HARK. Therefore, it was natural that HARK was applied to the system that recognizes orders for dishes from three persons at the same time, which had been used as a benchmark. It presently works on robots such as Robovie-R2 and HRP -2. As a variation of three-speaker simultaneous speech recognition, a referee robot that judges the winner of the rock-paper-scissors game played orally by three persons has been developed on Robovie-R2 [17]. Moreover, as a non-robot application, we have developed a system that visualizes the sounds that HARK localizes and separates based on real time or archived data. In presentation of sounds, the situation that “the sound is not detected” is often seen in some environments. We captured this problem as lack of auditory awareness (sound recognition). In order to improve the auditory awareness, we designed a three-dimensional sound environment visualization system that supports comprehension of sound environment and implemented in HARK [18, 19].