A 52 mW Full HD 160-Degree Object Viewpoint Recognition SoC With Visual Vocabulary Processor for Wearable Vision Applications

Abstract

A wearable 1920×1080 160-degree object viewpoint recognition SoC is realized on a 6.38mm 2 die with 65nm CMOS technology. This system focuses on enhancing the capability for wide viewpoint and long-distance recognition while reducing the computation of feature matching process. The recognition accuracy is improved from 29% to 94% under full HD resolution for a 50m-far traffic light compared with the performance under VGA (640×480). Object viewpoint prediction (OVP) supports 160-degree object viewpoint differences. 85% of power consumption and 75% of memory bandwidth are reduced via proposed visual vocabulary processor (VVP). 52mW power consumption with 25.9GOPS/mm 2 area efficiency is achieved.

An Energy-Efficient Matching Accelerator Using Matching Prediction for Mobile Object Recognition

“…One of the most challenging issues in object matching is its huge external memory bandwidth required in accessing object DB [1]. Vocabulary tree (VT) is well known for its efficient use of memory bandwidth by quantizing the key points of objects in the DB, which makes the entire DB fit in the on-chip memory [1,4].…”

“…Vocabulary tree (VT) is well known for its efficient use of memory bandwidth by quantizing the key points of objects in the DB, which makes the entire DB fit in the on-chip memory [1,4]. However, the VT-based object matching incurs considerable transactions to internal memory as it compares input object with the entire DB.…”

“…Recently, object recognition becomes wide spread in mobile multimedia applications and plays a key role in augmented reality and autonomous robots [1]. Object matching is one of the major parts in this object recognition process [1] and selects the closest item with input object from the object database (DB).…”

“…Object matching is one of the major parts in this object recognition process [1] and selects the closest item with input object from the object database (DB). It involves massive accesses to external memory containing the DB [1], and therefore, energy efficient design of this object matching is of the most importance in mobile vision applications [2].…”

“…Object matching is one of the major parts in this object recognition process [1] and selects the closest item with input object from the object database (DB). It involves massive accesses to external memory containing the DB [1], and therefore, energy efficient design of this object matching is of the most importance in mobile vision applications [2]. A variety of object matching algorithms have been studied, and recently, the vocabulary tree (VT) is gaining attentions for its efficient use of external memory bandwidth [3,4].…”