Improvement of odor recognition chip

Since a handheld odor-sensing system with high computation capability for a neural network is required for the field measurement, a digital LSI called an odor recognition chip with the neural network was developed In our circuit, the data are expressed as 1-bit data stream with delta-sigma modulation to reduce the amounts of circuit and wiring. The LVQ (Learning Vector Quantization) circuit was designed using that technique. The experimental result revealed that the chip recognized the odors almost correctly even if the sensor responses were remarkably changed due to the air turbulence.