Identification of Solid and Liquid Materials Using Acoustic Signals and Frequency-Graph Features.

Area of Science:

• Materials Science
• Acoustics
• Signal Processing
• Mobile Computing

Background:

• Material identification is crucial across industries but current methods are often contact-based, costly, and lack portability.
• Existing non-contact methods like Wi-Fi or radar lack easy integration into portable devices.

Purpose of the Study:

• To develop a novel, non-contact material identification system leveraging smartphone acoustics.
• To utilize the built-in microphone and speaker of smartphones as a transceiver for material analysis.

Main Methods:

• Acoustic signals are used to probe materials, creating distinct multipath profiles based on material properties.
• Channel Impulse Response (CIR) measurements were extracted, followed by image feature extraction (HOG, GLCM) from time-frequency domain graphs.
• An Error-Correcting Output Code (ECOC) learning method with majority voting was employed for material classification.

Main Results:

• A prototype system was built using three Android mobile phones.
• The system achieved average identification accuracies of 90% for solid materials and 97% for liquid materials.
• Successful material identification was demonstrated across varied multipath environments.

Conclusions:

• Smartphone-based acoustic material identification offers a portable and cost-effective solution.
• The distinct multipath profiles generated by acoustic signals serve as reliable material fingerprints.
• This technology has the potential for widespread application in various sectors requiring material identification.