PAM offers powerful front end processing capability for the system to talk to a speech recognizer from up to 10 feet away. The unit is designed for small to large rooms, from acoustically hard (kitchens, bathrooms) to acoustically soft (family rooms, bedrooms) surfaces in a home where the dynamic noise environment can be from extreme quiet to quiet loud.

This PAM module embeds in the Device as a front end processing, control, I/O, and communication engine. PAM is 3x 4 hardware and is fully integrated with front end acoustic algorithms including Automatic Gain Control, Voice Activity Detectors, Broadband Acoustic Echo Canceller, Speaker Identification, and Attention Word. Depending on the Device functionality and industrial design, appropriate speakers, microphones, display, and keys in an appropriate enclosure are required.

PAM Hardware Specifications
PAM Integrated Software/Algorithm Specifications
PAM Integration Documentation and Technical Specifications

Network Interface: RJ-45 Ethernet 10/100
USB: One Host and one Host or Device (optional)
Audio I/O: Lineout, LineIn, Headphone, Ext. Speaker
I/O: LED, key, control, IRDA
Display: On-board graphics controller
Processor: Up to 500 MHz DSP/RISC
Memory: 16/ 32 / 64 Mbytes SDRAM; 16/32 Mbytes flash
Power: 5V DC
Speech Input: Dual tuned microphone
Music/Speech Output: Dual Speakers
Interface to Custom Display Board: 4 wire SPI
Memory Expansion: (optional) Buffered bus for supporting mass storage and memory expansion option for PCMCIA and Compact Flash
Physical Size: 3x 4

Signal Normalization over multiple microphones
PA-Base communication drivers
Double Talk Detector
Broadband acoustic echo canceller
Background noise canceller
Attention Word Always listening
Customized front end acoustic models
Biometric identification
MP3, WMA, and other decoder (optional)
Windows CE/Linux OS option