Breast ultrasound (BUS) is an effective clinical modality for diagnosing breast abnormalities in women. Deep-learning techniques based on convolutional neural networks (CNN) have been widely used to analyze BUS images. However, the low quality of B-mode images owing to speckle noise and a lack of training datasets makes BUS analysis challenging in clinical applications. In this study, we proposed an end-to-end CNN framework for BUS analysis using multiple parametric images generated from radiofrequency (RF) signals. The entropy and phase images, which represent the microstructural and anatomical information, respectively, and the traditional B-mode images were used as parametric images in the time domain. In addition, the attenuation image, estimated from the frequency domain using RF signals, was used for the spectral features. Because one set of RF signals from one patient produced multiple images as CNN inputs, the proposed framework overcame the limitation of datasets in a broad sense of data augmentation while providing complementary information to compensate for the low quality of the B-mode images. The experimental results showed that the proposed architecture improved the classification accuracy and recall by 5.5% and 11.6%, respectively, compared with the traditional approach using only B-mode images. The proposed framework can be extended to various other parametric images in both the time and frequency domains using deep neural networks to improve its performance.

Various power estimation methods with high-level hardware power model, i.e. power emulation, had been proposed to overcome the slow speed of traditional power estimation in software environment. But precise speed comparison between gate-level simulation and high-level power emulation was conducted rarely. In this paper, we propose an architecture of power emulation which can compare precise speed gain against the traditional power simulation. Experimental results show that the average power error is 0.29%, the power emulation analyzes power 20,206 times faster than the traditional power simulation.

Breast ultrasound (BUS) is an effective clinical modality for diagnosing breast abnormalities in women. Deep-learning techniques based on convolutional neural networks (CNN) have been widely used to analyze BUS images. However, the low quality of B-mode images owing to speckle noise and a lack of training datasets makes BUS analysis challenging in clinical applications. In this study, we proposed an end-to-end CNN framework for BUS analysis using multiple parametric images generated from radiofrequency (RF) signals. The entropy and phase images, which represent the microstructural and anatomical information, respectively, and the traditional B-mode images were used as parametric images in the time domain. In addition, the attenuation image, estimated from the frequency domain using RF signals, was used for the spectral features. Because one set of RF signals from one patient produced multiple images as CNN inputs, the proposed framework overcame the limitation of datasets in a broad sense of data augmentation while providing complementary information to compensate for the low quality of the B-mode images. The experimental results showed that the proposed architecture improved the classification accuracy and recall by 5.5% and 11.6%, respectively, compared with the traditional approach using only B-mode images. The proposed framework can be extended to various other parametric images in both the time and frequency domains using deep neural networks to improve its performance.

Various power estimation methods with high-level hardware power model, i.e. power emulation, had been proposed to overcome the slow speed of traditional power estimation in software environment. But precise speed comparison between gate-level simulation and high-level power emulation was conducted rarely. In this paper, we propose an architecture of power emulation which can compare precise speed gain against the traditional power simulation. Experimental results show that the average power error is 0.29%, the power emulation analyzes power 20,206 times faster than the traditional power simulation.

전력 에뮬레이션을 통해 빠르고 정확한 소프트웨어 전력 분석이 가능함을 사례연구를 통해 보인다. 기존에 제안된 하드웨어 전력 모델인 전력계산기를 이용하여 사례연구를 수행하였다. FPGA에 구현된 전력계산기를 게이트수준 전력분석 환경과 동일한 조건에서 검증하기 위하여 PEU(power emulation unit)을 제안한다. 검증한 전력계산기를 이용하여 프로세서와 부동소수점 연산기를 포함하는 시스템의 전력 에뮬레이션을 수행하였다. 다수의 부동소수점 연산을 소프트웨어만으로 수행했을 때 보다 하드웨어 소프트웨어 파티셔닝을 통해 수행했을 때, 4배 이상의 성능향상과 58% 이상의 에너지 감소가 가능함을 전력 에뮬레이션을 이용해 보였다.

문턱 전압이 낮아지고, 칩의 집적도가 높아져서 온도가 증가함에 따라 누설 전력이 크게 증가하고 있다. 본 논문에서는 누설 전력을 상위수준에서 빠르게 분석할 수 있는 누설 전력 모델링 기술을 제안한다. 게이트 입력 값에 따라 누설전력량의 변함을 이용하여 입력의 정적 확률을 기반으로 누설 전력을 예측할 수 있음을 보인다. 또한 복수의 게이트로 이루어진 회로로 이를 확장하여 그 실용성을 증명한다. n-입력 회로의 경우, 2SUPn/SUP개의 계수를 이용하여 누설 전력 모델의 수식이 커지는 문제를 n+1개의 계수만을 이용하는 모델로 유도하는 과정을 설명한다. 또한 모델링시 생성해야 하는 학습 입력 벡터를 일정하게 만드는 법을 제안하여 학습 입력 벡터에 전력 모델의 정확도가 종속되지 않도록 하였다. 실험 결과 전력 모델의 정확도는 평균 98%, 속도 향상은 평균 218배를 보였다.

메모리 분할은 하나의 메모리가 아닌 두 개 또는 그 이상의 메모리를 이용하여 전력 소모를 줄이는 저전력 기법 중 하나이다. 본 논문에서는 SRAM 분할 시 발생하는 deselected power를 줄이기 위한 SRAM 분할 구조를 제안한다. Deselected power는 SRAM이 활성화되지 않았을 때, 주소나 데이터 입력 신호에서 변화가 발생하여 소모되는 동적 전력이다. 실험에서는 32KB 크기에 대해 다양한 크기의 SRAM 이용 시의 면적과 메모리 접근 패턴에 따른 전력 소모에 대한 분석을 진행한다. 그리고 제안하는 SRAM 분할 구조를 OpenRISC 프로세서의 캐시 데이터 메모리에 적용하고 CoreMark 벤치마크 소프트웨어 수행 시의 전력 소모를 비교한다. OpenRISC에 적용한 결과는 8개의 4KB SRAM 사용 시 면적은 27%가 증가하지만, 전력 소모는 명령어 캐시와 데이터 캐시의 데이터 메모리에서 각각 41%와 50%만 소모하여 전체 소모 전력도 23%가 감소하였다.

Energy efficiency is a very important requirement in designing a MAC protocol for wireless sensor networks using battery-operated sensor nodes. We proposed a new energy-efficient MAC protocol, RIX-MAC, based on asynchronous duty cycling and receiver-initiated scheme. In this article, we analyze the performance such as throughput, delay, and energy consumption of RIX-MAC with modeling and simulation. For mathematical analysis, we first use the Markov chain model and determine the transmission and state probabilities and set the equations to solve throughput and delay. We also calculate the energy consumption by separating a cycle period into TX and RX durations. Our analysis results are validated by comparing with the simulation results obtained by NS-2.