III. SENSOR SIGNAL ANALYSIS
In the measurement procedure, we register the differential signals of all pairs of sensors as a function of time and process
them in order to find the numerical features characterizing the differences in smells with the highest possible sensitivity. The measurement of each coffee specimen was registered 100 times (in a dynamical mode) within the measurement window. The registration of each sample needed 0.3 s, so the whole measurement of 100 coffee specimens lasted 30 s. The temporary pattern of sensor signals registered by our system will form the individual input for each classifier, responsible for the final recognition. Fig. 2 presents the differential signal patterns of 100 samples measured by four chosen sensors at three coffee preparations: pure arabica, pure robusta, and 50% mixture of robusta and arabica. They represent the dynamic response of the sensors (the change of the sensor resistance with respect to their original values converted into voltage) versus time. The blue (solid line) represents arabica, green (dash line) represents robusta, and red (dash-dotted line) represents the 50% mixture of robusta and arabica. As it is seen, the sensors react in
different ways to the presence of gas representing smells of different coffee types. The levels of sensor signals are changing a lot, depending on the type of sensor and the particular smell. The dynamic range of the differential sensor signals changes a lot for different sensors. In our investigation, it was extending from the minimum value equal to 0.0045 V for sensor 9 to the maximum range of 0.99 V for sensor 4. Some of the differential signals are negative, which means that the presence of gas in temporal state reduces the signal value of the sensor with respect to the signal of the reference array sensor. In further processing, we consider the temporary differential signal values of all sensors instead of their averages in the steady state, usually practiced in classical solutions of e-noses. The temporary pattern formed by the differential sensor signals will form the basis of the final recognition of smell. Such way of processing makes the system more flexible and able to make recognition on the basis of the measurement made in either steady or transient
state of the process.
Fig. 3 presents the average values of the original and normalized differential signals for all 12 sensors, calculated for 100 samples registered at the measurement of pure robusta, pure arabica, and 50% mixture of coffee robusta and arabica.
III. SENSOR SIGNAL ANALYSISIn the measurement procedure, we register the differential signals of all pairs of sensors as a function of time and processthem in order to find the numerical features characterizing the differences in smells with the highest possible sensitivity. The measurement of each coffee specimen was registered 100 times (in a dynamical mode) within the measurement window. The registration of each sample needed 0.3 s, so the whole measurement of 100 coffee specimens lasted 30 s. The temporary pattern of sensor signals registered by our system will form the individual input for each classifier, responsible for the final recognition. Fig. 2 presents the differential signal patterns of 100 samples measured by four chosen sensors at three coffee preparations: pure arabica, pure robusta, and 50% mixture of robusta and arabica. They represent the dynamic response of the sensors (the change of the sensor resistance with respect to their original values converted into voltage) versus time. The blue (solid line) represents arabica, green (dash line) represents robusta, and red (dash-dotted line) represents the 50% mixture of robusta and arabica. As it is seen, the sensors react indifferent ways to the presence of gas representing smells of different coffee types. The levels of sensor signals are changing a lot, depending on the type of sensor and the particular smell. The dynamic range of the differential sensor signals changes a lot for different sensors. In our investigation, it was extending from the minimum value equal to 0.0045 V for sensor 9 to the maximum range of 0.99 V for sensor 4. Some of the differential signals are negative, which means that the presence of gas in temporal state reduces the signal value of the sensor with respect to the signal of the reference array sensor. In further processing, we consider the temporary differential signal values of all sensors instead of their averages in the steady state, usually practiced in classical solutions of e-noses. The temporary pattern formed by the differential sensor signals will form the basis of the final recognition of smell. Such way of processing makes the system more flexible and able to make recognition on the basis of the measurement made in either steady or transientstate of the process.Fig. 3 presents the average values of the original and normalized differential signals for all 12 sensors, calculated for 100 samples registered at the measurement of pure robusta, pure arabica, and 50% mixture of coffee robusta and arabica.
การแปล กรุณารอสักครู่..
