Voice recognition appts. using dynamic comparison - uses multi-processor architecture to compare sound sample with stored patterns almost simultaneously

摘要

The appts. has two processing units (1,2). The sound signal input is sampled digitally. The first slice of the sampled signal is received by the first processor and processed to produce vectors of frequency analysis. The vectors are written to a FIFO memory unit (4). The second processor starts the process of dynamic comparison between the contents of the FIFO memory (4) and its local memory (21). The process now continues in parallel with no requirement to assemble the complete vector pattern of the sound input. USE/ADVANTAGE - Speech recognition with minimum time between input and response.