• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.461-467
• /
• 2020

In noisy and multi-speaker environments, the performance of speech recognition is unavoidably lower than in a clean environment. To improve speech recognition, in this paper, the signal of the speaker of interest is extracted from the mixed speech signals with multiple speakers. The VoiceFilter model is used to effectively separate overlapped speech signals. In this work, clustering by Probabilistic Linear Discriminant Analysis (PLDA) similarity score was employed to detect the speech signal of the interested speaker, which is used as the reference speaker to VoiceFilter-based separation. Therefore, by utilizing the speaker feature extracted from the detected speech by the proposed clustering method, this paper propose a speaker diarization system using only the mixed speech without an explicit reference speaker signal. We use phone-dataset consisting of two speakers to evaluate the performance of the speaker diarization system. Source to Distortion Ratio (SDR) of the operator (Rx) speech and customer speech (Tx) are 5.22 dB and -5.22 dB respectively before separation, and the results of the proposed separation system show 11.26 dB and 8.53 dB respectively.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.414-423
• /
• 2020

In this paper, we propose a Dilated Convolution Gate Linear Unit (DCGLU) to mitigate the lack of sparsity and small receptive field problems caused by the segmentation map extraction process in sound event detection with weak labels. In the advent of deep learning framework, segmentation map extraction approaches have shown improved performance in noisy environments. However, these methods are forced to maintain the size of the feature map to extract the segmentation map as the model would be constructed without a pooling operation. As a result, the performance of these methods is deteriorated with a lack of sparsity and a small receptive field. To mitigate these problems, we utilize GLU to control the flow of information and Dilated Convolutional Neural Networks (DCNNs) to increase the receptive field without additional learning parameters. For the performance evaluation, we employ a URBAN-SED and self-organized bird sound dataset. The relevant experiments show that our proposed DCGLU model outperforms over other baselines. In particular, our method is shown to exhibit robustness against nature sound noises with three Signal to Noise Ratio (SNR) levels (20 dB, 10 dB and 0 dB).

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.4
• /
• pp.83-90
• /
• 2010

In wireless sensor networks, the data transmitted from the sensor nodes are susceptible to corruption by errors which caused of noisy channels and other factors. In view of the severe energy constraint in Sensor Networks, it is important to use the error control scheme of the energy efficiently. In this paper, we presented RS (Reed-Solomon) codes in terms of their BER performance and power consumption. RS codes work by adding extra redundancy to the data. The encoded data can be stored or transmitted. It could have errors introduced, when the encoded data is recovered. The added redundancy allows a decoder to detect which parts of the received data is corrupted, and corrects them. The number of errors which are able to be corrected by RS code can determine by added redundancy. The results of experiment validate the performance of proposed method to provide high degree of reliability in low-power communication. We could predict the lifetime of RS codes which transmitted at 32 byte a 1 minutes. RS(15, 13), RS(31, 27), RS(63, 57), RS(127,115), and RS(255,239) can keep the days of 173.7, 169.1, 163.9, 150.7, and 149.7 respectively. The evaluation based on packet reception ratio (PRR) indicates that the RS(255,239) extends a sensor node's communication range by up about 3 miters.

• Geophysics and Geophysical Exploration
• /
• v.5 no.1
• /
• pp.23-32
• /
• 2002

Capacitively-coupled resistivity (CCR) system is known to be very useful where galvanic contact to earth is impossible, such as the area covered with thick ice, snow, concrete or asphalt. This system injects current non-galvanically, i.e., capacitively to earth through line antenna and measures potential difference in a same manner. We derived geometric factor for two types of antenna configuration and presented the method of processing and converting the data obtained with CCR system suitable to conventional resistivity inversion analysis. The CCR system, however, has limitations on use at conductive area or electrically noisy area since it is very difficult to inject sufficient current to earth with this system as with conventional resistivity system. This causes low SM ratio when acquiring data with CCR system and great care must be taken in acquiring data with this system. Additionally the uniform contact between line antennas and earth is also crucial factor to obtain good S/N ratio data. The CCR method, however, enables one to perform continuous profiling over a survey line by dragging entire system and thus will be useful in rapid investigation of conductivity distribution in shallow subsurface.

• Geophysics and Geophysical Exploration
• /
• v.2 no.4
• /
• pp.180-183
• /
• 1999

A proto-type seismic measurement system with a reference was developed to improve S/N (signal-to-noise ratio) of seismic data, especially in noisy urban areas. Two pairs of correlation measurements (the one for microphone and geophone, and another for electromagnetic (EM) loop and geophone) were carried out near Kimpo Airport and at Kimje. The spectrum analyses were also performed to investigate the correlation of two pairs of time series; one for microphone and geophone, and another for EM loop and geophone. The sound waves measured with the microphone and the geophone are highly correlated. However, differences in the reponses are readily identifiable across 200 Hz; in the vicinity of 100 Hz, the spectral energy for geophone is 20 dB higher than that for microphone, and at near 500 Hz, the spectral energy for microphone is 30 dB higher than that for geophone. Overall, the spectral energy appears concentrated on the frequency window below 600 Hz for geophone. It contrasts with the observation of dominant frequency at the range of above 200 Hz for microphone. The wave forms of EM noise (due to an ACDC inverter) measured with EM loop and geophone are consistently and highly correlated each other. The power spectrum of the EM noise for EM loop shows that the spectral energies at odd harmonic frequencies of 60 Hz are higher than those at even harmonic frequencies of 60 Hz. It is compared to the power spectrum for geophone; the spectral energies at odd harmonics are nearly same as those at even harmonic frequencies.

• Geophysics and Geophysical Exploration
• /
• v.12 no.4
• /
• pp.299-308
• /
• 2009

A magentotelluric (MT) survey at the Uiseong area has been performed for the site investigation of pilot scale $CO_2$ sequestration. The purpose of the MT survey is to delineate deeply extended fracture systems that can act as a leakage path of injected $CO_2$ Plume. Since the target area is extremely noisy in electromagentic sense, low frequency data below 1 Hz cannot be used for inversion. Two- and three-dimensional interpretation of the MT data showed a very clear conductive anomaly, which has the direction of $N55\sim65^{\circ}W$ and is extended roughly down to 1.6 km. It have the same direction with the strike-slip faults, the Gaeum and Geumcheon Faults. On the contrary, the eastern part of the survey area shows relatively homogeneous to the depth of 2 km though some small fractures at shallow depths can be found. Test drilling and high-definition borehole surveys should be followed at the eastern part of the survey area and hydraulic fracturing is required for injection of $CO_2$, because mean porosity of the sedimetary rock in the area is only 1.47%.

• Perspectives in Nursing Science
• /
• v.10 no.1
• /
• pp.12-23
• /
• 2013

Purpose: This study was conducted to investigate ICU nurses' perceptions of communication difficulties, the importance of and satisfaction with communication with doctors, other nurses, patients, and family, as well as to explore communication barrier with patient families. Methods: Investigators developed a 15-item communication perception questionnaire and 58-item communication barrier questionnaire. Communication barrier included 4 domains: nurses, family, environment, and patient condition. A total of 151 ICU nurses with a minimum of one year of ICU experience participated. Results: ICU patients ($3.38{\pm}0.73$) were the most difficult group to communicate with, followed by family ($3.32{\pm}0.72$), senior nurses ($3.25{\pm}0.74$), doctors ($3.21{\pm}0.68$), and nurse colleagues ($2.64{\pm}0.73$). Doctors ($4.61{\pm}0.53$) were the most important group to communicate with, followed by nurse colleagues ($4.52{\pm}0.54$), patients ($4.49{\pm}0.58$), senior nurses ($4.44{\pm}0.55$), and family ($4.43{\pm}0.61$). Satisfaction with communication was the highest with colleague nurses ($3.60{\pm}0.68$), then senior nurses ($3.37{\pm}0.74$), family ($3.18{\pm}0.71$), patients ($3.09{\pm}0.75$), and doctors ($3.06{\pm}0.83$).The total score of the communication barrier was $2.83{\pm}0.52$, where each domain was scored as follows: patient condition $3.13{\pm}0.74$, nurses $2.83{\pm}0.60$, environment $2.81{\pm}0.66$, and family $2.76{\pm}0.57$. The ICU nurses reported that communication was difficult due to 'sudden deterioration in the patient's condition', 'being too busy', 'a noisy environment', and 'information not being shared between family members.' Significant differences were noted by age, clinical experience, and marital status of nurse respondents. Conclusion: The findings indicated that development of a protocol on communication between nurses and doctors as well as development of an educational program on communication skills are necessary.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.5 s.305
• /
• pp.95-102
• /
• 2005

When images compressed with block-based compression techniques are transmitted over a noisy channel, unexpected block losses occur. In this paper, we present a post-processing-based block recovery scheme using Haar wavelet features. No consideration of the edge-direction, when recover the lost blocks, can cause block-blurring effects. The proposed directional recovery method in this paper is effective for the strong edge because exploit the varying neighboring blocks adaptively according to the edges and the directional information in the image. First, the adaptive selection of neighbor blocks is performed based on the energy of wavelet subbands (EWS) and difference of DC values (DDC). The lost blocks are recovered by the linear interpolation in the spatial domain using selected blocks. The method using only EWS performs well for horizontal and vertical edges, but not as well for diagonal edges. Conversely, only using DDC performs well diagonal edges with the exception of line- or roof-type edge profiles. Therefore, we combined EWS and DDC for better results. The proposed methods out performed the previous methods using fixed blocks.

• Phonetics and Speech Sciences
• /
• v.12 no.3
• /
• pp.65-71
• /
• 2020

Feature normalization is a method to reduce the effect of environmental mismatch between the training and test conditions through the normalization of statistical characteristics of acoustic feature parameters. It demonstrates excellent performance improvement in the traditional Gaussian mixture model-hidden Markov model (GMM-HMM)-based speech recognition system. However, in a deep neural network (DNN)-based speech recognition system, minimizing the effects of environmental mismatch does not necessarily lead to the best performance improvement. In this paper, we attribute the cause of this phenomenon to information loss due to excessive feature normalization. We investigate whether there is a feature normalization method that maximizes the speech recognition performance by properly reducing the impact of environmental mismatch, while preserving useful information for training acoustic models. To this end, we introduce the mean and exponentiated variance normalization (MEVN), which is a compromise between the mean normalization (MN) and the mean and variance normalization (MVN), and compare the performance of DNN-based speech recognition system in noisy and reverberant environments according to the degree of variance normalization. Experimental results reveal that a slight performance improvement is obtained with the MEVN over the MN and the MVN, depending on the degree of variance normalization.

• Journal of Navigation and Port Research
• /
• v.27 no.1
• /
• pp.97-102
• /
• 2003

Wire ropes are used in a myriad of various industrial applications such as elevator, mine hoist, construction machinery, lift, and suspension bridge. Especially, the wire rope of crane is important component to container transfer. If it happens wire rope failures during the operation, it may lead to safety accident, economic loss by productivity decline and so on. To solve this problem, we developed remote wire rope fault detecting system, and this system is consisted of 3 parts that portable fault detecting part, signal processing part and remote monitoring part. All detected signal has external noise or disturbance according to circumstances. So, we applied to discrete wavelet transform to extract a signal from noisy data. It is verified that the detecting system by de-noising has good efficiency for inspecting faults of wire ropes in service. As a result, by developing this system, container terminal could reduce expense because of extension fo wire ropes exchange period and could competitive power. Also, this system is possible to apply in several field such as elevator, lift and so on.