{"title":"DCT-based scheme to accelerate multimedia search in NAND Flash memories","authors":"Shruti Vyas, Aswin Sreedhar, S. Kundu","doi":"10.1109/SOCDC.2010.5682972","DOIUrl":null,"url":null,"abstract":"NAND Flash memories are gaining popularity due to their high density, robustness, low power consumption and low read times. Searching data in NAND flash memory is fast for small memory sizes but as the memory size increases, searches become painfully slow. With increased user content inclining towards large multimedia such as images, audio and video, there is a need for faster multimedia content searches (MCS). In this paper we develop a hardware based enhancement technique for fast multimedia content searches in NAND Flash memories. The central idea is to compress the multimedia data by applying Discrete Cosine Transform (DCT) and storing selected coefficients as signatures in the spare blocks of the memory. When a multimedia search request comes in, a signature of the search request is computed and only the signature blocks are compared for a match, thus making faster searches. DCT-based compression gives good results for text, audio, image and video files. If a small part of a multimedia data is given as search request, this technique returns all possible matches found in the set of files stored in the flash memory. Applications of this technique can be found in entertainment industry, music libraries, face recognition etc. Simulations are run for memories between size 2Gb to 16Gb. A speed-up of 450X in the search operation is achieved with this technique. The additional hardware has no performance impact on read or sequential writes of memory. The hardware overhead is estimated to be 0.03% of the total memory area.","PeriodicalId":380183,"journal":{"name":"2010 International SoC Design Conference","volume":"222 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2010-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2010 International SoC Design Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/SOCDC.2010.5682972","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
Abstract
NAND Flash memories are gaining popularity due to their high density, robustness, low power consumption and low read times. Searching data in NAND flash memory is fast for small memory sizes but as the memory size increases, searches become painfully slow. With increased user content inclining towards large multimedia such as images, audio and video, there is a need for faster multimedia content searches (MCS). In this paper we develop a hardware based enhancement technique for fast multimedia content searches in NAND Flash memories. The central idea is to compress the multimedia data by applying Discrete Cosine Transform (DCT) and storing selected coefficients as signatures in the spare blocks of the memory. When a multimedia search request comes in, a signature of the search request is computed and only the signature blocks are compared for a match, thus making faster searches. DCT-based compression gives good results for text, audio, image and video files. If a small part of a multimedia data is given as search request, this technique returns all possible matches found in the set of files stored in the flash memory. Applications of this technique can be found in entertainment industry, music libraries, face recognition etc. Simulations are run for memories between size 2Gb to 16Gb. A speed-up of 450X in the search operation is achieved with this technique. The additional hardware has no performance impact on read or sequential writes of memory. The hardware overhead is estimated to be 0.03% of the total memory area.