The need for human-machine interaction such as speech and gesture recognition has steadily grown in wearable devices. As applications become more intelligent such as facial emotion recognition, a variety of recognition algorithms has been developed and evolving. However, as the recognition algorithms become more complex, the more computation is required to perform the application in a limited battery capacity of wearable devices, which means that energy-efficiency is critical issue. In this thesis, one of the widely used recognition algorithm, artificial neural network (ANN), is selected as a target algorithm and its characteristic, inherent algorithmic fault tolerance (AFT), is adopted to improve energy-efficiency. To compute the recognition algorithm (ANN), Significant-driven iterative approximate multiplier (SDIAM) is utilized. Motivated by the fact that both an iteration of multiplication and the number of hidden nodes play key roles for a trade-off between recognition accuracy and energy consumption, these two parameters are optimized for a minimum of energy consumption of ANN, allowing acceptable recognition accuracy. The evaluation shows that the joint optimization between the iteration of multiplication and the number of hidden nodes save 70% of the energy consumption, compared with using precise computation, at the same recognition accuracy target for both handwritten and isolated spoken digit recognition. Furthermore, adopting SDIAM in training phase, the recognition accuracy is more improved, which leads to 87% and 75% lower energy consumption for handwritten and isolated spoken digit recognition. ⓒ 2016 DGIST
Table Of Contents
I. Introduction 1-- II. Recognition Applications on Wearable Devices 4-- 2.1. Trend on Recognition Applications 4-- 2.1.1. Vision-Based Applications 4-- 2.1.2. Sound-Based Application 5-- 2.1.3. Other Sensor-Based Application 6-- 2.2. Recognition Algorithms for Application 6-- 2.2.1. Artificial Neural Network (ANN) 7-- 2.2.2. Support Vector Machine (SVM) 8-- 2.2.3. Hidden Markov Model (HMM) 9-- 2.2.4. Deep Learning 10-- III. Artificial Neural Network 12-- 3.1 Introduction to ANN 12-- 3.2 Architecture and Feedforward Operation of ANN 12-- 3.3 Learning Algorithm 14-- 3.4 A demand for selecting the optimal number of hidden nodes 14-- IV. Energy-Efficient Hardware Accelerator for ANN 16-- 4.1 Characteristics of ANN to apply SDIAM 16-- 4.2 Significance-Driven Iterative Approximate Multiplier (SDIAM) 16-- 4.2.1. Architecture and Operation of SDIAM 16-- 4.2.2. Recognition Accuracy and Energy Consumptions 19-- 4.3. Joint Optimization of N and n 19-- 4.4 Training ANN with SDIAM 20-- V. Performance Evaluation 22-- 5.1. Evaluation Methodology 22-- 5.2. Evaluation 22-- VI. Related Works 30-- VII. Conclusion 32