In the Artificial Intelligence (AI) world, research works are so focusing on image processing which was very successful, as the famous saying "A picture is better than a thousand words", the technology of image process and face recognition has been developed to be advanced in many application areas. However, AI technologies should include all senses, vision, hearing, touch, taste, and smell, all five are the whole picture, Cloud-LED is a local company, they have study many years on smell, by sensors array with AI design, the smell of coffee or wine could be distinguished. The smelling implement into sensors involves the understanding of chemistry, computing, and electronics. By combining the knowledge of quantum chemistry, AI algorithm and doped metal oxide altogether.
A more challenging project for your student project is an electronic nose, you may need time to study and collect all kinds of gas sensors. Gas sensors simulate human odorant receptors and the circuit of gas sensors array functionally as an olfactory system by artificial intelligence algorithm. The sense of smell is one of the most important of human senses. The basic principles of the human olfactory system comprised of some 1,000 olfactory receptor types. These receptors are located on the olfactory receptor cells, they are in the upper part of the nasal epithelium and detect the inhaled odorant molecules. Each olfactory receptor cell possesses only one type of odorant receptor, and each receptor can detect a limited number of odorant substances. Our olfactory receptor cells are therefore highly specialized for a few odors. Receptor cells carrying the same type of receptor send their nerve processes to the brain, where the information from several olfactory receptors is combined, forming a pattern. Therefore, we can consciously experience the smell of a flower or some bad smell. In this invention, we use the mechanism in physiology apply to the electronic system by using gas sensors as the odor receptors, and circuit as a neural system, the numerical calculation algorithm implemented into software the memory of smell became the extraction of AI model's coefficients. A unique odor can trigger distinct memories that are the sensors array triggers the sensors array to the AI model's parameters. All their learned AI model's coefficients are saved in computer memory and become the "memory of smell", when the electronic nose smell a similar odor, it triggers the matching of AI model's coefficients and trigger the "smell" and "remember" what the gas is. This invention consists of circuit design, selection of sensors combination, and the sensors array could be in the form of an integrated circuit (IC), where the sensors array consists of thousands of gas sensors to constructed with sensing AI circuits. The invention, called "electronic nose", formally is a gas sensor functionally useful to work as odorant receptors and the circuit of gas sensors array, it functions as an olfactory system by artificial intelligence algorithm. It consists of functions as an electronic nose could help to smell unique gas combination, no matter good smell or bad smell. A large family of odorant receptors in the electronic nose build by gas sensing materials including tin oxide, indium oxide and the combination of both in a special ratio. Gas selecting elements, for example, platinum doped into tin oxide will select gas like hydrogen. Diversify materials for doping into tin oxide/indium oxide, and the host materials are not limited to the only tin oxide/indium oxide, it also applies in special ceramics. The sensors array works as odorant receptors are activated by an odorous substance, as an electric signal is triggered in the olfactory receptor cell and sent to the brain via nerve processes. The invented electronic nose work in an electronic circuit ways. Artificial intelligence electronic circuit collects data and model coefficient into a computer database, after the learning process, that is the artificial intelligence model has been trained, the effect as the sensitivity of individual olfactory receptor cells (gas sensor) to specific odorants. Most odors are composed of multiple odorant molecules, and each odorant molecule activates some odorant receptors. It works like a combinatorial code forming an odorant pattern, for example like many small pieces of glass mosaic combine to a beautiful picture. All these odorant receptors are implemented into sensors array in this invention. It is the ability of the number of gas sensors classified into less than 100 but the ability to recognize over 10,000 different odors. The invention of the electronic nose is a biomimetic of human receptor cells in the human body, the imagination of the biological system for engineering use also called it biomimicry. It is an electronic circuit design and artificial intelligence to mimic human odorant receptors, it has not consisted of genetic technology.
If you take this challenging problem, you may see we are going to use a series of gas sensors to make an electronic nose. If we could collect 16 different kinds of gas sensors, and collect signal from each sensor, the signal value from each sensor range from 0 to 1023, and put it into a tensor, simply a one-dimensional array now. [0, 0, 0, 0, 125, 123] If you see this, it seems there is no meaning to you, right? But if these numbers are gas sensors corresponding to a below chemical element, what will you say.
[B, C, N, O, H, S] It’s H(Hydrogen) and S(Sulfur) has a signal, the other has no signal, it seems that a gas smell related to chemical element hydrogen and sulfur, you may think about it is hydrogen sulfide gas. You see it!
So, you already see that if a series of gas sensors, each one has the sensitivity to some particular chemical elements, once the chemical composition of the gas exists related element, there is received signal from the corresponding gas sensor, then you may gas the combination. This is fit for artificial intelligence to gas since the AI model will rate the fitting coefficient to each element in the above list with some numerical numbers. This is the input tensor of the gas signal. In order to “smell” and distinguish it, the voltage of the sensor will charge from 5V to 3.3V for 16 different values. If we doing so, each “smell” could get 16 x 16 of signal values, they are our raw data input to a neural network to train the model. There will be a lot of information if the sensors take a “smell” acquiring data every minute. In a smart building, we need a lot of sensors to ensure the building is safe, all device is running smoothly, and energy saving. A message broker (by an edge processor) to collect data and send data, it is a “middle person” in a message transaction, that’s called a broker. In a building, effectively there are many systems, sensors’ data are collected to the cloud, after analysis, control signal sent to actuators including air-condition control fan, cool water flow, and pump. In a smart building, there are many energy collecting devices. please design your energy collecting devices (energy harvesting) including thermoelectric, solar power panel (photovoltaics), and piezoelectric, as a combined energy harvesting device to be used. How will it be?
If you are interested in the project of "smell science", you may be interested to study more in the area of quantum chemistry, ceramic semiconducting and computing, you can also contact firstname.lastname@example.org to bring the "Science of Smell" at your organization, if you like to have the STEM course book, please contact email@example.com, or send a message at www.stemedkit.com website. Thank you.