Abstract
Denoising of signals in an Internet-of-Things (IoT) network is
critically challenging owing to the diverse nature of the nodes
generating them, environments through which they travel, characteristics
of noise plaguing the signals and the applications they cater to. In
order to address the abovementioned challenges, we conceptualize a
generalized framework combining wavelet packet transform (WPT) and
energy correlation analysis. WPT decomposes both the low-frequency and
high-frequency components of the received signals in different time
scales and wavelet spaces. Noise components are identified, removed
through filtering and the signal components are predicted back after
filtering using inverse wavelet packet transform (IWPT). Next energy of
the reconstructed signal components are compared with that of the
original transmitted signal to modify the characteristics of the
decomposed signal components. Using the modified details, the signal
components are reconstructed back again and the noise components are
filtered out. This process is repeated until noise is completely
removed. Initial results suggest that, our proposed framework offers
improvement in error probability performance of a medium-scale IoT
network over traditional discrete wavelet transform (DWT) and WPT based
techniques by around 3 dB and 7 dB respectively.