Generation of ultrasonic and audible sound waves for the automatic classification of packaging waste in reverse vending machines

Reverse vending machines (RVMs) are essential for promoting waste sorting at the source by offering incentives for recycling. However, current RVMs, which primarily rely on expensive sensors such as barcode scanners and computer vision systems, face limitations including unreadable barcodes, high computational demands, and sensitivity to environmental conditions like lighting. This paper presents an alternative approach using acoustic sensors for waste classification. The proposed method consists of emitting ultrasonic and audible sound waves towards the recyclable object and, by analyzing the variations in the acoustic field, an artificial intelligence system classifies the material. For doing so, the system uses the ultrasonic and audible impulse response of each item, measured using the exponential sine sweep (ESS) technique. To validate this approach, a proof-of-concept has been developed and tested in a controlled environment using a scaled replica of a reverberation chamber, designed to achieve ideal acoustic conditions. Acoustic impulse responses have been captured using ESS emitted by an omnidirectional parametric loudspeaker (OPL), which generates both ultrasonic and audible sound waves via the parametric acoustic array (PAA) effect. This setup allows for simultaneous collection of ultrasonic and audible impulse responses for each recyclable item. The collected acoustic data has then been used to train classical machine learning and deep learning models to classify the introduced material, specifically plastic, glass, cardboard, and metallic cans. The results demonstrate a promising classification accuracy of more than 90%, highlighting the potential of this acoustic technology for waste sorting and supporting further research into its application in RVMs.