Young-Sun Yun

A segmental-feature HMM for continuous speech recognition based on a parametric trajectory model

In this paper, we propose a new acoustic model for characterizing segmental features and an algorithm based upon a general framework of hidden Markov models (HMMs). The segmental features are represented as a trajectory of observed vector sequences by a polynomial regression function. To obtain the polynomial trajectory from speech segments, we modify the design matrix to include transitional information for contiguous frames. We also propose methods for estimating the likelihood of a given segment and trajectory parameters. The observation probability of a given segment is represented as the relation between the segment likelihood and the estimation error of the trajectories. The estimation error of a trajectory is considered the weight of the likelihood of a given segment in a state. This weight represents the probability of how well the corresponding trajectory characterizes the segment. The proposed model can be regarded as a generalization of a conventional HMM and a parametric trajectory model. We conducted several experiments to establish the effectiveness of the proposed method and the characteristics of the segmental features. The recognition results on the TIMIT database demonstrate that the performance of segmental-feature HMM (SFHMM) is better than that of a conventional HMM.

A perspective on the IoT services through a multi-dimensional analysis

IoT (Internet of Things) technology connects network with intelligent things. The application range of this technology has been expanding through convergence with technologies from various fields. It has no certain limitation on application, rather than that it creates a new service while having convergence with existing systems and services. IoT services have variety of kinds that makes them hard to understand the precise features such as the scale, capacity and performance. There needs a categorization method for those services which is both inclusive and simplified to understand easily. This paper presents a new multi-dimensional classification method for IoT services. In our classification model, IoT services are classified as three dimensions; temporal, spatial, and relational criteria. The multi-dimensional classification model contributes to the high level understanding of IoT services.

Design and Implementation of μ-Webpage based on QR Code

QR(Quick Response) Code has been developed to provide greater storage capacity and more functionality compared to 1D bar codes. With the emergence of increasingly mobile devices equipped with cameras such as smart-phones and tablets, QR codes have become very popular and more important in mobile businesses. Typically, most QR codes are used as a URL link for redirecting users to webpages. However, the URL based QR codes are required to be connected over the internet and to be run a server. This can incur unnecessary traffics in the Internet. Furthermore, it is not suitable for the country lagging behind others in its network infrastructure. In this paper, we propose a server-less -webpage to provide mobile web services and be optimized for the capabilities and limitations of QR Code. We have implemented the -webpage in Android, and the results showed that the proposed mechanism can provide web-services without requiring extra servers or incurring mobile traffic data compared to the URL-based QR Codes.

Development of an O2O(offline to online) application case based on person wide web platform

As smartphones become more and more popular, O2O (Offline-to-Online) applications have been spotlighted. They provide users offline data using NFC, QR, and Beacon technologies. But, there are some problems such as the high cost of server maintenance and the private data spills since they are implemented on WWW technologies. This paper suggests PWW (Person Wide Web) platform that provides server-less service on smartphones. In PWW, a smartphone has a PWW browser which is composed of an anchor scanner, a web view, an API linkage library, and internal/cloud storages. When the PWW browser scans an anchor includes a URL, it opens a web document stored in the internal storage of the smartphone or its corresponding cloud. The web document contains the API linkage library to utilize the several devices like a camera, GPS, and so on in the smartphone. In order to show the usefulness of PWW platform, we described the example of handling private data such as business cards. The PWW platform could be applied to various area in real world by developing web-based mobile applications.

Deep neural networks based user interface detection for mobile applications using symbol marker

The UI storyboard is a design drawing for the development of mobile application. The designer uses a storyboard as a tool to help communication between the developer and the designer. However, the developer can make an unintended widget which is different from the intention of the designer. In this paper, we propose a DNN (Deep Convolutional Neural Network) based automatic UI detection method using symbol markers to improve the accuracy of UI identification. The symbol marker can be useful for UI detection and developing the UI Code.

Detection of GUI Elements on Sketch Images Using Object Detector Based on Deep Neural Networks

Graphical user interface (GUI) is very important to interact with software users. In many studies, therefore, they are trying to convert GUI elements (or widgets) to code or to describe formally its structure by help of domain knowledge or machine learning based algorithms. In this paper, we adopted object detection based on deep neural networks that finds GUI elements by integration of localization and classification. After the successfully detection of GUI components, we will describe the objects as the hierarchical structure and transform those to appropriate codes by synthetic or machine learning algorithms.

A Dynamic Duty Cycle Adjustment Mechanism for Reduced Latency in Industrial Plants

For environmental monitoring and risk identification of industrial plants, several monitoring systems using Wireless Sensor Networks (WSNs) have been developed. In this paper, we propose a dynamic duty cycle adjustment mechanism for reduced latency in industrial plants. The proposed method adjusts the duty cycle among predefined risk groups depending on the urgency of sensed data values. To demonstrate its efficacy, we analyze the expected transmission latency model and then discuss the characteristics in detail. We show that the proposed dynamic duty cycle mechanism is a more effective than a periodic mechanism by analyzing the expected latency of them in industrial plants where there are various types of sensory data with different levels of reliability.

Voice Conversion Between Synthesized Bilingual Voices Using Line Spectral Frequencies

Voice conversion is a technique that transforms the source speaker individuality to that of the target speaker. We propose the simple and intuitive voice conversion algorithm not using training data between different languages and it uses text-to-speech generated speech rather than recorded real voices. The suggested method reconstructed the voice after transforming line spectral frequencies (LSF) by formant space warping functions. The formant space is the space consisted of representative four monophthongs for each language. The warping functions are represented by piecewise linear equations using pairs of four formants at matched monophthongs. In this paper, we applied LSF to voice conversion because LSF are not overly sensitive to quantization noise and can be interpolated. From experimental results, LSF based voice conversion shows good results for ABX and MOS tests than the direct frequency warping approaches.

Design of tandem architecture using segmental trend features

This paper investigates the tandem architecture (TA) based on segmental features. The segmental feature based recognition system has been reported to show better results than the conventional feature based system in previous studies. In this paper we tried to merge the segmental feature with the tandem architecture which uses both hidden Markov models and neural networks. In general, segmental features can be separated into the trend and location. Since the trend means variation of segmental features and since it occupies a large portion of segmental features, the trend information was used as an independent or additional feature for the speech recognition system. We applied the trend information of segmental features to TA and used posterior probabilities, which are the output of the neural network, as inputs of the recognition system. Experiments were performed on Aurora2 database to examine the potentiality of the trend feature based TA. The results of our experiments verified that the proposed system outperforms the conventional system on very low SNR environments. These findings led us to conclude that the trend information on TA can be additionally used for the traditional MFCC features.

Trend tying in the segmental-feature HMM

We present a reduction method for the number of parameters in a segmental-feature HMM (SFHMM). If the SFHMM shows better results than the CHMM, the number of parameters is greater than that of the CHMM. Therefore, there is a need for a new approach that reduces the number of parameters. In general, the trajectory can be separated by the trend and location. Since the trend means the variation of segmental features and occupies a large portion of the SFHMM, if the trend is shared, the number of parameters of the SFHMM may be decreased. The proposed method shares the trend part of trajectories by quantization. The experiments are performed on the TIMIT corpus to examine the effectiveness of the trend tying. The experimental results show that its performance is the almost same as that of previous studies. To obtain better results with a small amount of parameters, the various conditions for the trajectory components must be considered.