An embedded electronic nose is developed including three modules, a sensor array, an A/D system, and a microprocessor. The sensor array composed of three gas sensors, TGS813,TGS825 and TGS880 produced by Figaro company, is used to collect sample signal. A MAX1270 chip produced by MAXIM company is used as A/D converter. The S3C2410 microprocessor produced by SAMSUNG company is used as CPU with the embedded operating system WINCE. The signal is simulated by a logarithm curve by designed software, and BP neural network is used to train the curve parameters and build the model to identify the samples. This system is used to identify Italian Ambrosi cheese with four different production dates, and the recognition accuracy reaches 100%. To identify pure milk in three different companies, the recognition accuracy reaches 100% too. The results show that the system has several advantages such as small size, light weight, easy to use, high recognition accuracy etc.