µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ÃÖÀû ÁÖÆļö ¿¬µ¿À» ÀÌ¿ëÇÑ ÀúÀü·Â °æ·® ÇàÀ§ÀÎ½Ä ¹æ¹ý |
¿µ¹®Á¦¸ñ(English Title) |
ÃÖÀû ÁÖÆļö ¿¬µ¿À» ÀÌ¿ëÇÑ ÀúÀü·Â °æ·® ÇàÀ§ÀÎ½Ä ¹æ¹ý |
ÀúÀÚ(Author) |
À̽ÂÇö
¹Ú±â¼º
È«ÁöÇý
Çѿ뱸
ÀÌ¿µ±¸
Seunghyun Lee
Kisung Park
Jihye Hong
Yongkoo Han
Young-Koo Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 28 NO. 01 PP. 0137 ~ 0153 (2012. 04) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù ¸ð¹ÙÀÏ ±â±â¸¦ ÀÌ¿ëÇÑ ÇàÀ§ÀÎ½Ä ±â¼úÀÌ ¿¬±¸µÇ°í ÀÖ´Ù. ¸ð¹ÙÀÏ ±â±â¸¦ »ç¿ëÇÏ¿© ½Ç¿Ü ȯ°æ¿¡¼ ´Ù¾çÇÑ »óȲ Á¤º¸·ÎºÎÅÍ ÀνÄÇÑ ÇàÀ§µéÀº ÇコÄÉ¾î ¾ÖÇø®ÄÉÀ̼ǿ¡ À¯¿ëÇÏ°Ô È°¿ëµÉ ¼ö ÀÖ´Ù. ±×·¯³ª ±âÁ¸ÀÇ ¸ð¹ÙÀÏ ±â±â¸¦ ÀÌ¿ëÇÑ ÇàÀ§ÀÎ½Ä ±â¹ýµéÀº Á¦ÇÑµÈ ¹èÅ͸® ¼ö¸í°ú ³·Àº ÄÄÇ»Æà ÆÄ¿ö ¹®Á¦¸¦ °¡Áö°í ÀÖ´Ù. º» ³í¹®Àº °¢ ÇàÀ§¸¦ ÃÖÀûÀÇ ¼¾¼ ÁÖÆļö¿¡¼ ÀνÄÇÏ¿© ¼¾¼ ¼Òºñ Àü·ÂÀ» ÁÙÀÌ´Â ÀúÀü·Â °æ·® ÇàÀ§ÀÎ½Ä ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. ÇàÀ§ÀÎ½Ä Á¤È®µµ´Â ÃÖÀû À©µµ¿ì »çÀÌÁî ¹× ¿À¹ö·¡ÇÎ ºñÀ²À» °¢ ÁÖÆļöº°·Î ¿¬µ¿ÇÔÀ¸·Î½á À¯ÁöÇÑ´Ù. ¶ÇÇÑ Æ¯Â¡ ÃßÃâ °úÁ¤ÀÇ Áߺ¹ ¿¬»êÀ» Á¦°ÅÇÏ¿© ¸ð¹ÙÀÏ ±â±âÀÇ ¼Òºñ Àü·ÂÀ» ÁÙÀÌ´Â °æ·®È ¹æ¾ÈÀ» ÇÔ²² Á¦½ÃÇÑ´Ù. ½ÇÇèÀ» ÅëÇÏ¿© Á¦¾ÈÇÏ´Â ±â¹ýÀÌ ±âÁ¸ÀÇ ¹æ¹ý°ú ºñ±³ÇÏ¿© ºñ½ÁÇÑ ÇàÀ§ÀÎ½Ä Á¤È®µµ¿¡¼ ÃÖ´ë 23%ÀÇ Àü·Â ¼Òºñ·®À» ÁÙÀÓÀ» º¸ÀδÙ.
|
¿µ¹®³»¿ë (English Abstract) |
Recently, activity recognition using mobile devices has been studied. Activities recognized from various outdoor context data can help various healthcare applications. However, previous activity recognition methods using mobile devices are suffering from limited computing power and short battery lifespan. In this paper, we propose a low-power lightweight activity recognition technique. It reduces power consumption of sensors by recognizing activities with their optimal frequencies rather than using a fixed frequency. Moreover, high recognition accuracy is achieved through adjusting window sizes and overlapping ratios according to each frequency. We also propose a lightweight method that removes redundant operations in a feature extraction step for reducing power consumption in the mobile device. Empirical results of the proposed method have shown 23% power consumption reduction compared with the existing approaches.
|
Å°¿öµå(Keyword) |
ÇàÀ§ÀνÄ
¼¾¼
ÀúÀü·Â
°æ·®È
¸ð¹ÙÀÏ ±â±â
Activity recognition
Sensor
Lightweight
Low-pow
Low-power
Mobile device
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|