Hitachi Makes Robot Good Listener w/ 14 Mics
Hitachi Ltd exhibited its second humanoid robot, "EMIEW2," at Hitachi uValue Convention 2008, the company's private show that took place from July 17 to 18, 2008, at Tokyo International Forum.
This is the second exhibition following the one in May 2008 at the Science Museum in Kitanomaru, Tokyo, the company said.
The EMIEW2 features (1) a light-weight (13kg) body, (2) the ability to move on wheels as well as on two legs and (3) the 14 microphones mounted in the head that enhance its voice recognition capabilities to the extent that it can distinguish voices even when three people are speaking at the same time.
(1) The weight of 13kg is much lighter than 70kg of the "EMIEW," which is the first version of the robot and was exhibited at the Aichi Expo in 2005.
"It was designed to be lighter than 15kg, which is considered to be the weight that any office worker can lift," Hitachi's spokesperson said.
The robot is 80cm tall, a height for looking over desk tops.
(2) The EMIEW2 has a wheel at the bottom of each leg and balances itself to move on flat floors just like the Segway, a two-wheeled vehicle ridden while standing. Basically, traveling on wheels is more efficient than walking. Therefore, the concept was based on "traveling on wheels" since the stage of developing the first version, Hitachi said.
However, the EMIEW2 features a walking function, which the EMIEW did not have, so that it can handle differences in floor level of a few centimeters, which wheels cannot.
(3) The voice recognition function is a combination of sound source separation and sound pattern recognition. In the sound source separation, sounds coming from directions other than the direction from which the voice of the person speaking to the robot comes from are ignored as noise. As a result, the robot can recognize voices spoken as far as 2m away, the company said.
In the sound pattern recognition, sound data is transmitted by using IEEE802.11a wireless LAN technology to a server machine, where the recognition processing is carried out. Then, the results are transmitted to the robot.
The processeing is handled by the robot when instant response is required, but the processing for conversation with humans, which requires a response speed of 100ms or more, is designed to be carried out by the server machine, Hitachi said.
The voice recognition function, too, employs a number of technologies developed by Hitachi.
"The robot can distinguish voices even when three people are talking to it at the same time," Hitachi said.
The company is not planning to commercialize the robot at this stage but said, "The price would be lower than that of a luxury car, if it were to be released."