Data fusion in universal domain using dual semantic code

In this work, the novel concept of data fusion is put forward, where any two or more signals are associated under the proposed framework. Data fusion aims at addressing the limitation of interchangeability in the conventional methods such as data embedding and metadata. First, the signals are mapped to the universal domain, which is a domain where signal of any underlying features can be represented. Then, the signals are universally parsed Abdul Karim and Wong (2014) [18] and encoded losslessly by DSC (Dual Semantic Code), where each codeword of DSC can accommodate two pieces of data simultaneously. Hence, data fusion is achieved by mapping a segment of each signal to different parts of a DSC codeword using two proposed coding schemes, namely, the basic and partial coding schemes. In the basic coding scheme, all segments of the largest signal (in terms of size for storage) are mapped to the DSC codewords. However, in the partial coding scheme, only segments of high probability of occurrences are mapped to the DSC codewords. The proposed coding schemes are universally applicable to any signal, such as image, video, audio and text. Experimental results suggest that the basic coding scheme achieves, on average, a fusion bit-rate of 0.640 bpb (bits per bit). On the other hand, the partial coding scheme achieves, on average, a fusion bit-rate of 0.060 bpb in the file size preserving mode. The fusion bit-rate can also be traded for compression ratio.