• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.56-60
• /
• 2023

기존의 음성인식은 청장년 층에 초점이 맞추어져 있었으나, 최근 고령화가 가속되면서 노인 음성에 대한 연구 필요성이 증대되고 있다. 그러나 노인 음성 데이터셋은 청장년 음성 데이터셋에 비해서는 아직까지 충분히 확보되지 못하고 있다. 본 연구에서는 부족한 노인 음성 데이터셋 확보에 기여하고자 희소한 노인 데이터셋을 증강할 수 있는 방법론에 대해 연구하였다. 이를 위해 노인 음성 특징(feature)을 분석하였으며, '주파수'와 '발화 속도' 특징을 일반 성인 음성에 합성하여 데이터를 증강하였다. 이후 Whisper small 모델을 파인 튜닝한 뒤 노인 음성에 대한 CER(Character Error Rate)를 구하였고, 기존 노인 데이터셋에 증강한 데이터셋을 함께 사용하는 것이 가장 효과적임을 밝혀내었다.

• The Journal of the Acoustical Society of Korea
• /
• v.36 no.5
• /
• pp.338-344
• /
• 2017

In this paper, an interpolation method of HRTF (Head-Related Transfer Function) is proposed for small-sized measurement data set, especially. The proposed algorithm is based on acoustic modeling of HRTFs, and the algorithm tries to interpolate the HRTFs via estimation the model coefficients. However, the estimation of the model coefficients is hard if there is lack of measurement points, so the algorithm solves the problem by a data augmentation using the VBAP (Vector Based Amplitude Panning). Therefore, the proposed algorithm consists of two steps, which are data augmentation step based on VBAP and model coefficients estimation step by least squares method. The proposed algorithm was evaluated by a simulation with a measured data from CIPIC (Center for Image Processing and Integrated Computing) HRTF database, and the simulation results show that the proposed algorithm reduces mean-squared error by 1.5 dB ~ 4 dB than the conventional algorithms.

• Journal of Auto-vehicle Safety Association
• /
• v.13 no.2
• /
• pp.35-41
• /
• 2021

In this study, improved path tracking control algorithm based on pure pursuit algorithm is newly proposed by using improved lane detection algorithm through real time post-processing with interpolation methodology. Since the original pure pursuit works well only at speeds below 20 km/h, the look-ahead distance is implemented as a sigmoid function to work well at an average speed of 45 km/h to improve tracking performance. In addition, a smoothing filter was added to reduce the steering angle vibration of the original algorithm, and the stability of the steering angle was improved. The post-processing algorithm presented has implemented more robust lane recognition system using real-time pre/post processing method with deep learning and estimated interpolation. Real time processing is more cost-effective than the method using lots of computing resources and building abundant datasets for improving the performance of deep learning networks. Therefore, this paper also presents improved lane detection performance by using the final results with naive computer vision codes and pre/post processing. Firstly, the pre-processing was newly designed for real-time processing and robust recognition performance of augmentation. Secondly, the post-processing was designed to detect lanes by receiving the segmentation results based on the estimated interpolation in consideration of the properties of the continuous lanes. Consequently, experimental results by utilizing driving guidance line information from processing parts show that the improved lane detection algorithm is effective to minimize the lateral offset error in the diverse maneuvering roads.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2007.12a
• /
• pp.98-99
• /
• 2007

It is necessary to use satellite radio navigation system as well as satellite radio navigation augmentation system such as differential Global Positioning System to achieve the positioning accuracy and reliability requested by International Maritime Organization in port and coastal area. Especially, position accuracy of DGPS user is effected by accuracy of pseudorange correction broadcasted from DGPS reference station. This paper shows pseudorange correction calculation algorithm adopting a non-common error estimation filter in order to improve accuracy of pseudorange correction. Finally, this paper verifies that the pseudorange correction calculated by adopting a non-common error estimation filter satisfies performance specifications of RTCM.

• Journal of Advanced Navigation Technology
• /
• v.25 no.3
• /
• pp.194-202
• /
• 2021

Ionosphere, one of the largest error sources, can pose potentially harmful threat to single-frequency GNSS (global navigation satellite system) user even after applying ionospheric corrections to their GNSS measurements. To quantitatively assess ionospheric impacts on the satellite navigation-based applications using simulation, the standard deviation of residual ionospheric errors is needed. Thus, in this paper, we determine conservative statistical quantity that covers typical residual ionospheric errors for nominal days. Extensive data-processing computes TEC (total electron content) estimates from GNSS measurements collected from the Korean reference station networks. We use Klobuchar model as a correction to calculate residual ionospheric errors from TEC (total electron content) estimate. Finally, an exponential delay model for residual ionospheric errors is presented as a function of local time and satellite elevation angle.

• The Journal of Engineering Geology
• /
• v.32 no.2
• /
• pp.221-239
• /
• 2022

Unknown geotechnical characteristics are key challenges in the design of piles for the plant, civil and building works. Although the N-values which were read through the standard penetration test are important, those N-values of the whole area are not likely acquired in common practice. In this study, the N-value is predicted by means of regression analysis with artificial intelligence (AI). Big data is important to improve learning performance of AI, so circular augmentation method is applied to build up the big data at the current study. The optimal model was chosen among applied AI algorithms, such as artificial neural network, decision tree and auto machine learning. To select optimal model among the above three AI algorithms is to minimize the margin of error. To evaluate the method, actual data and predicted data of six performed projects in Poland, Indonesia and Malaysia were compared. As a result of this study, the AI prediction of this method is proven to be reliable. Therefore, it is realized that the geotechnical characteristics of non-boring points were predictable and the optimal arrangement of structure could be achieved utilizing three dimensional N-value distribution map.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.6
• /
• pp.521-531
• /
• 2020

QZSS (Quasi-Zenith Satellite System) provides the CLAS (Centimeter Level Augmentation Service) through the satellite's L6 band. CLAS provides correction messages called C-SSR (Compact - State Space Representation) for GPS (Global Positioning System), Galileo and QZSS. In this study, CLAS messages were received by using the AsteRx4 of Septentrio which is a GPS receiver capable of receiving L6 bands, and the messages were decoded to acquire C-SSR. In addition, Multi-GNSS (Global Navigation Satellite System) Code-PPP (Precise Point Positioning) was developed to compensate for GNSS errors by using C-SSR to pseudo-range measurements of GPS, Galileo and QZSS. And non-linear least squares estimation was used to estimate the three-dimensional position of the receiver and the receiver time errors of the GNSS constellations. To evaluate the accuracy of the algorithms developed, static positioning was performed on TSK2 (Tsukuba), one of the IGS (International GNSS Service) sites, and kinematic positioning was performed while driving around the Ina River in Kawanishi. As a result, for the static positioning, the mean RMSE (Root Mean Square Error) for all data sets was 0.35 m in the horizontal direction ad 0.57 m in the vertical direction. And for the kinematic positioning, the accuracy was approximately 0.82 m in horizontal direction and 3.56 m in vertical direction compared o the RTK-FIX values of VRS.

• Korean Chemical Engineering Research
• /
• v.61 no.1
• /
• pp.123-141
• /
• 2023

A sufficient amount of data with quality is needed for training artificial neural networks (ANNs). However, developing ANN models with a small amount of data often appears in engineering fields. This paper presented an ANN model to improve prediction performance of the ammonia emission amount with 83 data. The ammonia emission rate included eleven inputs and two outputs (maximum ammonia loss, N_max and time to reach half of N_max, K_m). Categorical input variables were transformed into multi-dimensional equal-distance variables, and 13 data were added into 66 training data using a generative adversarial network. Hyperparameters (number of layers, number of neurons, and activation function) of ANN were optimized using Gaussian process. Using 17 test data, the previous ANN model (Lim et al., 2007) showed the mean absolute error (MAE) of K_m and N_max to 0.0668 and 0.1860, respectively. The present ANN outperformed the previous model, reducing MAE by 38% and 56%.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.6
• /
• pp.1312-1317
• /
• 2013

Pseudolite is a contraction of the term "pseudo-satellite", used to refer to something that is not a satellite which performs a function commonly in the domain of satellites. Pseudolite are most often small transceivers that are used to create a local, ground-based GPS alternative. Pseudo-range measurement of pseudolite has around 300m range error, when time synchronization error of $1{\mu}sec$ occurs. Therefore the time synchronization methods play an important part in navigation augmentation using pseudolite. This paper proposes three clock synchronization methods that are installation method of pseudolite station, method using KRISS-UTC and method using PRN code phase difference for pseudolite. The simulation platform structure is presented for evaluating proposed clock synchronization performance.

• Journal of Advanced Navigation Technology
• /
• v.23 no.3
• /
• pp.237-244
• /
• 2019

Since SBAS provides users with GNSS orbit, clock, and ionospheric corrections and integrity, the more precise positioning is possible. As the SBAS service area is expanded due to the development of the SBAS and the installation of the additional ground stations, there is a region where two or more SBAS messages can be received. However, the research on multi-constellation SBAS selection method has not carried out. In this study, we compared the result of positioning accuracy after applying the SBAS correction selected by using WAAS priority, EGNOS priority, or error covariance comparison method to LEO satellites in the regions where WAAS and EGNOS signals are transmitted simultaneously. When using WAAS priority method, 3D orbit error is smallest at 2.57 m. The covariance comparison method is outperform at the center of the overlap region far from each WAAS and EGNOS stations. In the eastern region near the EGNOS stations, the 3D orbit errors using EGNOS priority method is 8% smaller than the errors using the WAAS priority method.