| dc.description.abstract | Non-volatile memories are becoming the dominant type of storage devices in modern computers because of their fast speed, physical robustness and high data density. However, there still exist many challenges, such as the data reliability issues due to noise. An important example is the memristor, which uses programmable resistance to store data. Memristor memories use the crossbar architecture and suffer from the sneak-path problem: when a memristor cell of high resistance is read, it can be mistakenly read as a low-resistance cell due to low-resistance sneak-paths in the crossbar that are parallel to the cell. In this work, we study new ways to correct errors using the inherent redundancy in stored data (called Natural Redundancy), and combine them with conventional error-correcting codes. In particular, we define a Huffman encoding for the English language based on a repository of books. In addition, we study data stored using convolutional codes and use natural redundancy to verify if decoded codewords are valid or invalid. We present statistics over the Viterbi Algorithm and its ability to decode convolutional codewords, then discuss Yen's Algorithm, an augmentation of the Viterbi Algorithm. Finally, we present an efficient algorithm to search for a list of the most likely codewords, and choose a codeword that meets the criteria of both natural redundancy and the ECC as the decoding solution. We find that this algorithm is no more powerful than Yen's Algorithm in terms of decoding noisy convolutional codewords, but does present some interesting ideas for further exploration across multiple fields of study. | en |
