The small voice interaction module and PaPeRo-mini (left) and PaPeRo-mini (right)
The small voice interaction module and PaPeRo-mini (left) and PaPeRo-mini (right)
[Click to enlarge image]

NEC has developed an inexpensive business-card-sized voice interaction module that can be mounted on various instruments including in-vehicles terminals, mobile phones, information appliances, kiosk terminals, and robots. Downsizing of this module was achieved by implementing the NEC's technology such as voice recognition/voice synthesis, scenario control, facial recognition, and CG display into the application processor made by NEC Electronics Corporation. Employing this module results in eliminating the need for specific design for each instrument, therefore improving the development efficiency.

The voice interaction module was implemented with features such as voice recognition, voice synthesis, noise tolerant signal processing, scenario engine (such as programming the motion according to the situation for a robot), sound source direction finding, facial recognition, computer graphics display, network communication, and RT middleware collaboration. These features are handled by NEC Electronics' application processor "MP Series." In the development, the developed board for embedded devices "platformOViA" was used. The application processor is integrated with four cores; three ARM9s (192 MHz operation) and one DSP core "SPX-K6" (192 MHz operation).

The outside dimensions of the voice interaction module are 55 x 100 x 32 mm. The voice interaction module is comprised of three printed boards; a main board, an extension board and an audio board. In addition to the application processor, the main board has 128 MB of main memory, 64 MB of flash memory, along with other components. The extension board has a CompactFlash card slot, as well as 128 MB of memory as extension memory. Two two-channel microphone inputs are mounted on the main board for audio input and a 16-channel microphone input is mounted on the audio-board. For audio output, a two-channel line output and a two-channel speaker output are also provided. For graphic images, there is one camera input and for image output, an NTSC composite video, a liquid crystal panel display and an analog RGB are provided.

NEC made a small robot, "PaPeRo-mini," which employs this module, as a prototype and confirmed the potency of the module by testing various functions including voice recognition/voice synthesis/interaction, sound source direction finding, facial recognition and behavior control. In addition, according to types and features of instruments, a scenario engine achieves integrated control, middleware collaboration for robots, and collaboration with external installments through network communications, by making a choice of above-mentioned functions.

The outside dimension of PaPeRo-mini is 250 x 170 x 179 mm. This robot is smaller than a robot named PaPeRo that was prototyped and showed to the public by NEC. The possible functions of PaPeRo-mini are fewer than those of PaPeRo. The possible functions of PaPeRo-mini may be limited," says NEC Corporate Communications Division.

This "Built-in small voice interaction module" achievement was realized as part of the commissioned projects of the New Energy and Industrial Technology Development Organization (NEDO), "21st Century Robot Challenge Program" and "Next generation robot common infrastructure development project" (FY2005-2007). In these projects, this module will be built in to five robots, with the intention to demonstrate the features and performances of the module during FY2007. These five robots are made and will be studied by National Institute of Advanced Industrial Science and Technology, Yaskawa Electric Corporation, Mitsubishi Heavy Industries, LTD., Tokyo Institute of Technology/Tokyo Metropolitan University, and NEC.