Wavelet Transform and MFCC-Based Detection of Electromagnetic Stress in in selected indoor plants: A Machine Learning Approach

Abstract

Electromagnetic stress in six indoor plant species was detected using wavelet transform and Mel frequency cepstral coefficients (MFCC) with a machine learning approach. The study was conducted in Bariyatu, Ranchi, India during 2022–2023. A completely randomized design with six replicates per species was employed. Electromagnetic stress was applied at three levels: control (0 µT), low (50 µT), and high (200 µT). The selected species were Dieffenbachia seguine, Sansevieria trifasciata, Chlorophytum comosum, Dracaena fragrans, Bryophyllum pinnatum, and Aloe x principis. Five physiological parameters were measured: chlorophyll content, leaf temperature, relative water content, electrolyte leakage, and MFCC peak frequency. Data were analyzed using one way ANOVA in CropStat software. Results showed significant (p < 0.001) electromagnetic stress effects across all species and parameters. Chlorophyll content decreased from 45.6 to 26.4 SPAD units. Leaf temperature increased from 23.9 to 32.0°C. Relative water content declined from 90.2 to 60.8%. Electrolyte leakage rose from 0.31 to 1.02 mS/cm. MFCC peak frequency shifted from 138.2 to 284.5 Hz. Aloe x principis exhibited the highest stress sensitivity while Sansevieria trifasciata showed maximum tolerance. Wavelet transform and MFCC features effectively discriminated electromagnetic stress levels. The machine learning approach achieved accurate classification of stress severity. This research demonstrated that electromagnetic stress significantly alters biophysical and spectral signatures in indoor plants.