• The Journal of the Acoustical Society of Korea
• /
• v.26 no.6
• /
• pp.243-249
• /
• 2007

We propose an effective voice-based gender identification using group delay(GD) Generally, features for speech recognition are composed of magnitude information rather than phase information. In our approach, we address a difference between male and female for GD which is a derivative of the Fourier transform phase. Also, we propose a novel way to incorporate the features fusion scheme based on a combination of GD and magnitude information such as mel-frequency cepstral coefficients(MFCC), linear predictive coding (LPC) coefficients, reflection coefficients and formant. The experimental results indicate that GD is effective in discriminating gender and the performance is significantly improved when the proposed feature fusion technique is applied.

• Journal of KIISE:Software and Applications
• /
• v.34 no.3
• /
• pp.226-234
• /
• 2007

In this paper, we develop a speech recognition system using a throat microphone. The use of this kind of microphone minimizes the impact of environmental noise. However, because of the absence of high frequencies and the partially loss of formant frequencies, previous systems developed with those devices have shown a lower recognition rate than systems which use standard microphone signals. This problem has led to researchers using throat microphone signals as supplementary data sources supporting standard microphone signals. In this paper, we present a high performance ASR system which we developed using only a throat microphone by taking advantage of Korean Phonological Feature Theory and a detailed throat signal analysis. Analyzing the spectrum and the result of FFT of the throat microphone signal, we find that the conventional MFCC feature vector that uses a critical pass filter does not characterize the throat microphone signals well. We also describe the conditions of the feature extraction algorithm which make it best suited for throat microphone signal analysis. The conditions involve (1) a sensitive band-pass filter and (2) use of feature vector which is suitable for voice/non-voice classification. We experimentally show that the ZCPA algorithm designed to meet these conditions improves the recognizer's performance by approximately 16%. And we find that an additional noise-canceling algorithm such as RAST A results in 2% more performance improvement.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.495-505
• /
• 2019

Typical algorithms of deep learning include CNN(Convolutional Neural Networks), which are mainly used for image recognition, and RNN(Recurrent Neural Networks), which are used mainly for speech recognition and natural language processing. Among them, CNN is able to learn from filters that generate feature maps with algorithms that automatically learn features from data, making it mainstream with excellent performance in image recognition. Since then, various algorithms such as R-CNN and others have appeared in object detection to improve performance of CNN, and algorithms such as YOLO(You Only Look Once) and SSD(Single Shot Multi-box Detector) have been proposed recently. However, since these deep learning-based detection algorithms determine the success of the detection in the still images, stable object tracking and detection in the video requires separate tracking capabilities. Therefore, this paper proposes a method of combining Kalman filters into deep learning-based detection networks for improved object tracking and detection performance in the video. The detection network used YOLO v2, which is capable of real-time processing, and the proposed method resulted in 7.7% IoU performance improvement over the existing YOLO v2 network and 20 fps processing speed in FHD images.

• The Journal of the Korea Contents Association
• /
• v.22 no.2
• /
• pp.10-21
• /
• 2022

In order to improve formal presentation attitudes such as presentation of job interviews and presentation of project results at the company, there are few automated methods other than observation by colleagues or professors. In previous studies, it was reported that the speaker's stable speech and gaze processing affect the delivery power in the presentation. Also, there are studies that show that proper feedback on one's presentation has the effect of increasing the presenter's ability to present. In this paper, considering the positive aspects of correction, we developed a program that intelligently corrects the wrong presentation habits and attitudes of college students through facial analysis of videos and analyzed the proposed program's performance. The proposed program was developed through web-based verification of the use of redundant words and facial recognition and textualization of the presentation contents. To this end, an artificial intelligence model for classification was developed, and after extracting the video object, facial feature points were recognized based on the coordinates. Then, using 4000 facial data, the performance of the algorithm in this paper was compared and analyzed with the case of facial recognition using a Teachable Machine. Use the program to help presenters by correcting their presentation attitude.

• Science of Emotion and Sensibility
• /
• v.21 no.1
• /
• pp.59-70
• /
• 2018

With the breakthrough of speech recognition technology, conversational agents have become pervasive through smartphones and smart speakers. The recognition accuracy of speech recognition technology has developed to the level of human beings, but it still shows limitations on understanding the underlying meaning or intention of words, or understanding long conversation. Accordingly, the users experience various errors when interacting with the conversational agents, which may negatively affect the user experience. In addition, in the case of smart speakers with a voice as the main interface, the lack of feedback on system and transparency was reported as the main issue when the users using. Therefore, there is a strong need for research on how users can better understand the capability of the conversational agents and mitigate negative emotions in error situations. In this study, we applied social strategies, "forewarning" and "apology", to conversational agent and investigated how these strategies affect users' perceptions of the agent in breakdown situations. For the study, we created a series of demo videos of a user interacting with a conversational agent. After watching the demo videos, the participants were asked to evaluate how they liked and trusted the agent through an online survey. A total of 104 respondents were analyzed and found to be contrary to our expectation based on the literature study. The result showed that forewarning gave a negative impression to the user, especially the reliability of the agent. Also, apology in a breakdown situation did not affect the users' perceptions. In the following in-depth interviews, participants explained that they perceived the smart speaker as a machine rather than a human-like object, and for this reason, the social strategies did not work. These results show that the social strategies should be applied according to the perceptions that user has toward agents.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.1-19
• /
• 2019

Artificial intelligence (AI) is one of the main driving forces leading the Fourth Industrial Revolution. The technologies associated with AI have already shown superior abilities that are equal to or better than people in many fields including image and speech recognition. Particularly, many efforts have been actively given to identify the current technology trends and analyze development directions of it, because AI technologies can be utilized in a wide range of fields including medical, financial, manufacturing, service, and education fields. Major platforms that can develop complex AI algorithms for learning, reasoning, and recognition have been open to the public as open source projects. As a result, technologies and services that utilize them have increased rapidly. It has been confirmed as one of the major reasons for the fast development of AI technologies. Additionally, the spread of the technology is greatly in debt to open source software, developed by major global companies, supporting natural language recognition, speech recognition, and image recognition. Therefore, this study aimed to identify the practical trend of AI technology development by analyzing OSS projects associated with AI, which have been developed by the online collaboration of many parties. This study searched and collected a list of major projects related to AI, which were generated from 2000 to July 2018 on Github. This study confirmed the development trends of major technologies in detail by applying text mining technique targeting topic information, which indicates the characteristics of the collected projects and technical fields. The results of the analysis showed that the number of software development projects by year was less than 100 projects per year until 2013. However, it increased to 229 projects in 2014 and 597 projects in 2015. Particularly, the number of open source projects related to AI increased rapidly in 2016 (2,559 OSS projects). It was confirmed that the number of projects initiated in 2017 was 14,213, which is almost four-folds of the number of total projects generated from 2009 to 2016 (3,555 projects). The number of projects initiated from Jan to Jul 2018 was 8,737. The development trend of AI-related technologies was evaluated by dividing the study period into three phases. The appearance frequency of topics indicate the technology trends of AI-related OSS projects. The results showed that the natural language processing technology has continued to be at the top in all years. It implied that OSS had been developed continuously. Until 2015, Python, C ++, and Java, programming languages, were listed as the top ten frequently appeared topics. However, after 2016, programming languages other than Python disappeared from the top ten topics. Instead of them, platforms supporting the development of AI algorithms, such as TensorFlow and Keras, are showing high appearance frequency. Additionally, reinforcement learning algorithms and convolutional neural networks, which have been used in various fields, were frequently appeared topics. The results of topic network analysis showed that the most important topics of degree centrality were similar to those of appearance frequency. The main difference was that visualization and medical imaging topics were found at the top of the list, although they were not in the top of the list from 2009 to 2012. The results indicated that OSS was developed in the medical field in order to utilize the AI technology. Moreover, although the computer vision was in the top 10 of the appearance frequency list from 2013 to 2015, they were not in the top 10 of the degree centrality. The topics at the top of the degree centrality list were similar to those at the top of the appearance frequency list. It was found that the ranks of the composite neural network and reinforcement learning were changed slightly. The trend of technology development was examined using the appearance frequency of topics and degree centrality. The results showed that machine learning revealed the highest frequency and the highest degree centrality in all years. Moreover, it is noteworthy that, although the deep learning topic showed a low frequency and a low degree centrality between 2009 and 2012, their ranks abruptly increased between 2013 and 2015. It was confirmed that in recent years both technologies had high appearance frequency and degree centrality. TensorFlow first appeared during the phase of 2013-2015, and the appearance frequency and degree centrality of it soared between 2016 and 2018 to be at the top of the lists after deep learning, python. Computer vision and reinforcement learning did not show an abrupt increase or decrease, and they had relatively low appearance frequency and degree centrality compared with the above-mentioned topics. Based on these analysis results, it is possible to identify the fields in which AI technologies are actively developed. The results of this study can be used as a baseline dataset for more empirical analysis on future technology trends that can be converged.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.149-156
• /
• 2012

Support vector machines (SVMs) are well known for their pattern recognition capability, but proper care should be taken to alleviate their inherent implementation cost resulting from high computational intensity and memory requirement, especially in embedded systems where only limited resources are available. Since the memory requirement determined by the dimensionality and the number of support vectors is generally too high for a cache in embedded systems to accomodate, frequent accesses to the main memory occur inevitably whenever the cache is not able to provide requested data to the processor. These frequent accesses to the main memory result in overall performance degradation and increased energy consumption because a memory access typically takes longer and consumes more energy than a cache access or a register access. In this paper, we propose a technique that reduces the number of main memory accesses by optimizing the data access pattern of the SVM-based classifier in such a way that the temporal locality of the accesses increases, fully utilizing data loaded into the processor chip. With experiments, we confirm the enhancement made by the proposed technique in terms of the number of memory accesses, overall execution time, and energy consumption.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.5
• /
• pp.895-907
• /
• 2020

With the recent increasing influence of multimedia content other than the text-based content, services that help to process information in content brings us great convenience. These services' representative features are searching and masking the sensitive data. It is not difficult to find the solutions that provide searching and masking function for text information and image. However, even though we recognize the necessity of the technology for searching and masking a part of the audio data, it is not easy to find the solution because of the difficulty of the technology. In this study, we propose web application that provides searching and masking functions for audio data using audio partitioning method. While we are achieving the research goal, we evaluated several speech to text conversion APIs to choose a proper API for our purpose and developed regular expressions for searching sensitive information. Lastly we evaluated the accuracy of the developed searching and masking feature. The contribution of this work is in design and implementation of searching and masking a sensitive information from the audio data by the various functionality proving experiments.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.1
• /
• pp.108-113
• /
• 2022

Speech recognition in noise situation is an important skill for effective communication. Hearing In Noise Test (HINT) has been suggested as a clinical tool to evaluate these aspects. However, this tool has not been used widely in domestic clinics. In this study, psychophysical aspects of HINT and burdens in clinical application were analyzed to improve the applicability of the tool. The difficulty in understanding speech in the elderly population is due to hearing loss based on aging of peripheral and central auditory pathways. As typical clinical cases, HINT scores for young and elderly listeners (20s vs 70s) were compared. Four conditions of HINT test were Quiet (Q), Noise Front (NF), Noise Right (NR), and Noise Left (NL). Quantitative scores showed that the elderly listener required more Signal to Noris Ratio (SNR) values than the younger counterpart in noisy situations. Although both showed Binaural Masking Level Difference (BMLD) effect, the strength was smaller in the elder. However, the age-matched normalized data were not established in detail for clinical application. Confirmed usefulness of HINT and the related improvement in clinical measuring procedure were suggested.

• Phonetics and Speech Sciences
• /
• v.14 no.3
• /
• pp.77-86
• /
• 2022

In this study, we reported the validation results of the Korean Affective Voice Database (KAV DB), an affective voice database available for scientific and clinical use, comprising a total of 113 validated affective voice stimuli. The KAV DB includes audio-recordings of two actors (one male and one female), each uttering 10 semantically neutral sentences with the intention to convey six different affective states (happiness, anger, fear, sadness, surprise, and neutral). The database was organized into three separate voice stimulus sets in order to validate the KAV DB. Participants rated the stimuli on six rating scales corresponding to the six targeted affective states by using a 100 horizontal visual analog scale. The KAV DB showed high internal consistency for voice stimuli (Cronbach's α=.847). The database had high sensitivity (mean=82.8%) and specificity (mean=83.8%). The KAV DB is expected to be useful for both academic research and clinical purposes in the field of communication disorders. The KAV DB is available for download at https://kav-db.notion.site/KAV-DB-75 39a36abe2e414ebf4a50d80436b41a.