Kwang-Seok Hong

Person authentication using face, teeth and voice modalities for mobile device security

In this paper, we propose an enhanced multimodal personal authentication system for mobile device security. The proposed approach fuses information obtained from face, teeth and voice modalities to improve performance. To integrate three modalities, we employ various fusion techniques such as the weighted-summation rule, K-NN, Fisher and Gaussian classifiers, and we then evaluate the authentication performance of the proposed system. The performance is evaluated on a database consisting of 1000 biometric traits that correspond to the face, teeth and voice modalities of 50 persons, i.e., 20 biometric traits per individual, in which these biometric traits are simultaneously collected by a smart-phone device. The experiment results integrating the three modalities showed the error rates of 1.64%, 4.70%, 3.06% and 1.98% for the weighted-summation rule, K-NN, Fisher and Gaussian classifier, respectively, and that the weight-summation rule outperformed the other classification approaches. In contrast, the error rates regarding a single modality were 5.09%, 7.75% and 8.98% for face, teeth, and voice modalities, respectively. From these results, we confirmed that the proposed method achieved a significant performance improvement over the methods using a single modality, and the results showed that the proposed method was very effective through various fusion experiments.

Multimodal biometric authentication using teeth image and voice in mobile environment

Mobile devices such as smart-phone, PDA and mobile-phone are vulnerable to theft and loss due to their small size and the characteristics of the environments in which they are used. A simple and convenient authentication system is required to protect private information stored in the mobile device. Therefore, we propose a new multimodal biometric authentication approach using teeth image and voice as biometric traits in this paper. The individual matching scores obtained from the teeth image and voice are combined using a weighted-summation operation, and the fused-score is utilized to classify an unknown user into the acceptance or rejection. The proposed method is evaluated using 1000 teeth images and voices, in which these are collected by smart-phone, i.e., one mobile device for 50 subjects. In the experiment results, the proposed method has an EER of 2.13%, and we demonstrate the effectiveness of the proposed method.

Hand gesture recognition system using fuzzy algorithm and RDBMS for post PC

In this paper, we implement hand gesture recognition system using union of fuzzy algorithm and Relational Database Management System (hereafter, RDBMS) module for Post PC (the embedded-ubiquitous environment using blue-tooth module, embedded i.MX21 board and note-book computer for smart gate). The learning and recognition model due to the RDBMS is used with input variable of fuzzy algorithm (fuzzy max-min composition), and recognize users dynamic gesture through efficient and rational fuzzy reasoning process. The proposed gesture recognition interface consists of three modules: 1) gesture input module that processes motion of dynamic hand to input data, 2) RDBMS module to segment significant gestures from inputted data, and 3) fuzzy max-min recognition module to recognize significant gesture of continuous, dynamic gestures and extensity of recognition. Experimental result shows the average recognition rate of 98.2% for significant dynamic gestures.

Personalized smart TV program recommender based on collaborative filtering and a novel similarity method

The viewing set-based method has difficulties ensuring that a user will enjoy recommended programs, and the model-based collaborative filtering method contains system-side real-time recommendation problems because most recent ratings cannot be applied in the recommendations and it has increased calculating costs due to the training process. In this paper, we propose a personalized program recommender for smart TVs using memory-based collaborative filtering with a novel similarity method that is robust to cold-start conditions and faster than the often-used, existing similarity method. The proposed method can improve the recommendation performance of electronic program guides and recommender applications for smart TVs. We determined the prediction accuracy of the ratings under various conditions in order to evaluate the proposed method. As a result, we confirmed that the proposed method is effective for cold-start conditions.

Game interface using hand gesture recognition

In this paper, we propose a real-time hand gesture recognition system based on difference image entropy using a stereo camera. Existing systems use hand detection primarily with some type of marker. Our system, however, uses a real-time hand image recognition system. In the detection step, we implement a depth map using a sum of absolute differences based on the acquired right-left image using a stereo camera. This system detects a foreground object and perceives it as a hand. The hand gesture recognition system uses the difference image entropy of the input image and the average image. To evaluate the performance of the proposed technology, we implement a recognition experiment using the hand gesture 240 database. The proposed method shows an average recognition rate of 85%. Using the proposed method, we implement a Chinese chess game based on hand gesture recognition.

Finger gesture-based mobile user interface using a rear-facing camera

We propose finger gesture-based user interaction on mobile devices, focusing on mobile phones equipped with rear-facing cameras. A user interacts with a mobile application by using finger gestures with the hand that holds the mobile phone, which is detected by video taken with the rear-facing camera. A preprocessing method uses skin-color segmentation, morphological operation and skeletonization for finger tracking. Our system generates mouse events and other commands reacting to finger gestures.

Teeth recognition based on multiple attempts in mobile device

Most traditional biometric approaches generally utilize a single image for personal identification. However, these approaches sometimes failed to recognize users in practical environment due to false-detected or undetected subject. Therefore, this paper proposes a novel recognition approach based on multiple frame images that are implemented in mobile devices. The aim of this paper is to improve the recognition accuracy and to reduce computational complexity through multiple attempts. Here, multiple attempts denote that multiple frame images are used in time of recognition procedure. Among sequential frame images, an adequate subject, i.e., teeth image, is chosen by subject selection module which is operated based on differential image entropy. The selected subject is then utilized as a biometric trait of traditional recognition algorithms including PCA, LDA, and EHMM. The performance evaluation of proposed method is performed using two teeth databases constructed by a mobile device. Through experimental results, we confirm that the proposed method exhibits improved recognition accuracy of about 3.6-4.8%, and offers the advantage of lower computational complexity than traditional biometric approaches.

An implementation of an FPGA-based embedded gesture recognizer using a data glove

A gesture recognizer based on a desktop PC, which uses existing wire/wireless communication modules, has several restrictions such as space limitations, movement limitations, and change in recognition capacity depending on the change in the background lighting conditions when obtaining a users meaningful gesture data from images. This paper proposes an embedded gesture recognizer that uses a data glove in order to solve these problems. The proposed embedded FPGA (field-programmable gate array)-based gesture recognizer comprises an input module, a recognition module, and a display module. The input module receives the data that is transmitted from a data glove through a UART. The recognition module determines whether one set of data is accurate by performing data calculations with a checksum function after receiving the input data and comparing it to the header byte. This module also analyzes the data from 17 distinct gestures and constructs recognition models, and then it extracts the hand gesture data and compares it to the recognition models to see if the gestures match any of the 17 models. The recognition module then transmits the recognition result to the display module. The display module displays the recognition result on an LCD screen. A data glove manufactured by 5DT was used to obtain the gesture inputs. The FPGA was the XC3S1000FG676 (Xilinx Inc.) and it was designed using VHDL. The experimental results showed a 94% average recognition rate when using the FPGA-based embedded gesture recognizer and the data glove.

Speech Emotion Recognition Using Spectral Entropy

This paper proposes a Gaussian Mixture Model (GMM)---based speech emotion recognition methods using four feature parameters; 1) Fast Fourier Transform(FFT) spectral entropy, 2) delta FFT spectral entropy, 3) Mel-frequency Filter Bank (MFB) spectral entropy, 4) delta MFB spectral entropy. In addition, we use four emotions in a speech database including anger, sadness, happiness, and neutrality. We perform speech emotion recognition experiments using each pre-defined emotion and gender. The experimental results show that the proposed emotion recognition using FFT spectral-based entropy and MFB spectral-based entropy performs better than existing emotion recognition based on GMM using energy, Zero Crossing Rate (ZCR), Linear Prediction Coefficient (LPC), and pitch parameters.

Performance Evaluation of Teeth Image Recognition System Based on Difference Image Entropy

In this paper, we propose improved Difference Image Entropy (herein after, DIE)-based teeth recognition system and input image selection method. The DIE estimation module computes the DIE coefficient reflecting histogram levels have peak positions from -255 to +255, after obtains gray scaled-difference image from input teeth image and average teeth image on random-collected reference images. For performance evaluation of DIE-based teeth image recognition system including in-put image selection method, we implemented individual teeth images-based teeth image recognition system using K-NN with PCA and 2D-DCT-based EHMM pattern recognition algorithms, and then they are coupled with suggested DIE-based in-put image selection method. After that we inspect availability and validity of the application by various experiments using DIE threshold values from 6.9 to 7.3 for suitable teeth image selection. In experimental results, the suggested teeth recognition system shows approximately from 3% to 15% improved recognition performance than traditional image-based biometric algorithms / methods.