Arduino-based Sound Acquisition System Using Fast Fourier Transform Algorithm

The farmers and vendors in the Philippines classify coconuts through knocking manually either by bare hands or knife as 'mala-uhog', 'mala-kanin' and 'mala-tenga'. The said technique is well-known but yet carries no scientific proof. Thus, the proponents conducted this study to develop a knocking device that automatically classify coconuts by characterizing its maturity according to meat thickness and peak frequency using Fast Fourier Transform (FFT) algorithm. Sixty coconuts were sampled; each being initially categorized by a `mangangatok' to the stage it belongs using the conventional method. Each were knocked using a knife at a constant distance and then opened right after for the measurement of its meat thickness using a Vernier Caliper. The recorded sounds were then analyzed using FFT Algorithm in OCTAVE. Statistical analysis show that each of the three stages has high correlation between 0.869 and 0.897 in terms of meat thickness versus peak frequency. Results showed that `malauhog' coconuts have peak frequency range of about 200 to 400 Hz, 400 to 600 Hz for `malakanin' and 600 to 800 Hz for `malatenga'. From these results, a standalone Arduino-based sound acquisition knocking device was developed. The Sound acquisition system is designed to be portable. Its functionality was tested by sampling 119 random coconuts. Results showed that out of 119 there were 10 errors made by the device, exhibiting 91.6% accuracy. The device has capabilities of classifying maturity stages of coconuts which provides efficiency for vendors and compatibility for consumers with no knowledge in the currently known method.