An electronic nose is a sensor that identifies a number of odors by measuring the relative concentrations of different gas molecules in the air. The applications are numerous: air quality control, agriculture, food and pharmaceutical industries, etc. The specificity of an electronic nose, compared to other gas sensors, is that it does not include gas separation devices. Discrimination is achieved by using the cross specificities of several sensors, associated with a dedicated signal processing. Capacitive Micromachined Ultrasound Transducer (CMUT) are MEMS initially dedicated to the transmission and reception of ultrasonic acoustic waves. A set of transducers can also be used to characterize gases on a gravimetric principle. This technology has been very popular during the last decade for the development of electronic noses, especially because it can be easily coupled with surface functionalities. Thus, a cMUT transducer with a dedicated chemical functionalization forms the basic unit of an electronic nose. The nose may consist of several tens of transducers, each functionalized with a particular chemistry. Thanks to the European Research Council (ERC) funding obtained by Elisa Vianello to develop silicon circuits inspired by the insect nervous system, we wish to go further by integrating in the same low-power system the cMUT sensors, a near-sensor electronics, a signal processing and a neuromorphic classification circuit. The objective is to propose solutions to obtain a robust, very low power and very compact cMUTs based artificial nose system. The work will concern the acquisition of a solid database as a prerequisite to the design of the system, the adoption of a Machine Learning approach to develop the signal processing and classification, the design of an ASIC circuit to demonstrate key functions and the exploration of different computational paradigms for a robust estimation of the odors in presence (Bayesian network, neural network, or others).