Identification of household electrical appliances

The aim of this work is to propose a new identification system for domestic electrical appliances. Our first contribution is to propose an identification system based on the use of statistical parameters of harmonics and the application of the KNN classifier combined with the voting rule method. The results obtained show that the extraction of 500 parameters, based on the estimation of the statistical mean and standard deviation, combined with KNN classification and the voting rule strategy, gives the best CR Classification Rate of 94.97%. Our second contribution, is to reduce dimensionality by using a compact parameter representation (called DWE) that is based on estimating the mean and standard deviation of the energy calculated at each dyadic decomposition level of the wavelet analysis. Two descriptors called LWE and WCC are also extracted from this analysis by applying respectively the logarithm of the Total Energy and the discrete cosine transform. The results show that the WCC descriptor gives a maximum CR of 98.13%.

Read more