• Proceedings of the KWS Conference
• /
• 1996.10a
• /
• pp.139-141
• /
• 1996

Among various welding parameters, the welding current which is inversely proportional to the tip-to-workpiece distance in GMAW is an essential parameter to monitor the GMAW process of horizontal fillet joints. For the case of weld defect such as overlap in horizontal fillet welding, therefore, the signal processing for process monitoring or automatic seam tracking should be modified by considering the weld pool surface geometry including the corresponding weld defect. In other words, the adequate signal processing algorithm is indispensible to improve the performance of the arc sensor. However, arc sensor algorithm already developed usually focus on weld seam tracing but do not considering the weld qualities. In this paper, various experiments were carried out to investigate the tendencies of the weld defects when weaving motion is added, and the experimental method based on 2$^n$ factorial design was proposed for deriving the mathematical model between the leg length and the various welding conditions. Moreover, a signal processing method based on the artificial neural network(Adaptive Resonance Theory) was proposed far discriminating the current signal of sound weld beads from that of weld beads with overlap. Finally, the algorithm for weld seam tracking combined with the mathematical modeling and the signal processing method was carried out to track the weld line in conjunction with the improvement of the weld qualities. The reliability of the proposed algorithms were evaluated through various experiments, which showed that the proposed algorithms could be effectively used for arc welding automation.

• Journal of the Korea Society of Computer and Information
• /
• v.21 no.12
• /
• pp.131-137
• /
• 2016

In this paper, the peak finding algorithms corresponding to the Autocorrelation Function (ACF), which are widely exploited for detecting the pitch of voiced signal, are proposed. According to various researchers, it is well known fact that the estimation of fundamental frequency (F0) in speech signal is not only very important task but quite difficult mission. The proposed algorithms, presented in this paper, are implemented by using many characteristics - such as monotonic increasing function - of ACF function. Thus, the proposed algorithms may be able to estimate both reliable and correct the fundamental frequency as long as the autocorrelation function of speech signal is accurate. Since the proposed algorithms may reduce the computational complexity it can be applied to the real-time processing. The speech data, is composed of Korean emotion expressed words, is used for evaluation of their performance. The pitches are measured to compare the performance of proposed algorithms.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.4
• /
• pp.857-866
• /
• 2009

In this paper, a noise cancellation-method using microphone array for digital hearing aids is proposed. The microphone array is located around the ear of a dummy. Speech sound is generated from the forward speaker positioned in the front of the dummy and noise sound is generated from the backward speaker. The speech and noise are mixed in the air space and entered into the microphones. VAD(voice activity detector) and ANC(adaptive noise cancellation) methods were used to eliminate noise in the sound of the microphones. 10 two-syllable words and 4 sentences were used for speech signals. Babble and car interior noise were used for noise signals. The performance of the proposed algorithm was evaluated by SNR(signal-to-noise ratio) and PESQ-MOS(perceptual evaluation of speech quality-mean opinion score). In babble noise condition, SNR was improved as much as $7.963{\pm}1.3620dB\;and\;3.968{\pm}0.6659dB$ for words and sentences respectively. In the case of car interior noise, SNR was improved as $10.512{\pm}2.0665dB\;and\;6.000{\pm}1.7642dB$ for words and sentences respectively. PESQ-MOS of the babble noise was improved as much as $0.1722{\pm}0.0861$ score for words and $0.083{\pm}0.0417$ score for sentences. And PESQ-MOS of the car interior noise was improved as $0.2661{\pm}0.0335$ score and $0.040{\pm}0.0201$ score for words and sentences respectively. It is verified that the proposed algorithm has a good performance in noise cancellation of microphone array for digital hearing aids.

• Phonetics and Speech Sciences
• /
• v.5 no.1
• /
• pp.99-108
• /
• 2013

Previous laboratory studies have shown that prosodic structures are encoded in the modulations of phonetic patterns of speech including suprasegmental as well as segmental features. In particular, effects of prosodic context on duration and intensity of syllables and words have been widely reported. Drawing on prosodically annotated large-scale speech data from the Buckeye corpus of conversational speech of American English, the current study attempted to examine whether and how prosodic prominence and phrase boundary of everyday conversational speech, as determined by a large group of ordinary listeners, are related to the phonetic realization of duration and intensity. The results showed that the patterns of word durations and intensities are influenced by prosodic structure. Closer examinations revealed, however, that the effects of prosodic prominence are not the same as those of prosodic phrase boundary. With regard to intensity measures, the results revealed the systematic changes in the patterns of overall RMS intensity near prosodic phrase boundary but the prominence effects are restricted to the nucleus. In terms of duration measures, both prosodic prominence and phrase boundary are the most closely related to the lengthening of the nucleus. Yet, prosodic prominence is more closely related to the lengthening of the onset while phrase boundary lengthens the coda duration more. The findings from the current study suggest that the phonetic realizations of prosodic prominence are different from those of prosodic phrase boundary, and speakers signal different prosodic structures through deliberate modulations of the internal phonetic structure of words and listeners attend to such phonetic variations.

• The KIPS Transactions:PartB
• /
• v.15B no.6
• /
• pp.595-602
• /
• 2008

The focused word of each sentence is a help in recognizing and understanding spoken Korean. To find the method of focused word spotting at spoken speech signal, we made an analysis of the average and variance of Fundamental Frequency and the average energy extracted from a focused word and the other words in a sentence by experiments with the speech data from 100 spoken sentences. The result showed that focused words have either higher relative average F0 or higher relative variances of F0 than other words. Our findings are to make a contribution to getting prosodic characteristics of spoken Korean and keyword extraction based on natural language processing.

• Journal of Broadcast Engineering
• /
• v.20 no.4
• /
• pp.568-579
• /
• 2015

Recently, mobile devices with touch display panel are widely used. Accuracy and reaction speed of touch signal are very important in touch devices. Therefore, we need to develop an effective algorithm to reduce touch noise quickly and accurately. This paper proposes a touch noise reduction algorithm using Kalman filtering in consideration of signal motion. First, a specific pre-emphasis processing is applied to an input signal so as to maximize the effect of Kalman filtering. In other words, a pure signal in the touch signal increases but noise in the touch signal decreases. Next, motion of the signal is detected. Motion estimation is performed only if motion is detected. If we detect motion by using the only neighborhood of the signal, we can reduce about 75% of the computation in comparison with examining the entire area. Finally, Kalman filtering using the previous state of current signal is performed. Experimental results show that the proposed algorithm suppresses touch noise sufficiently without degradation of the pure signal

• The Journal of the Acoustical Society of Korea
• /
• v.8 no.5
• /
• pp.70-77
• /
• 1989

In this paper, Korean language recognition using the phoneme is studied. The experiment is carried out by dividing 545 isolated words into phonemes. Using linear prediction coefficients the recognition rate of consonants, vowels, and end-consonants are $87.3(\%), 91.0(\%), 91.7(\%)$, respectively. Recognition rate of isolated words combined with the phonemes is $71.4(\%)$. Itakura-saito distortion measure is used to phoneme segmentation and phoneme recognition.

• Journal of KIISE:Software and Applications
• /
• v.32 no.8
• /
• pp.742-751
• /
• 2005

This paper proposes a novel text image watermarking algorithm based on word classification. The words are classified into K classes using simple features. Several adjacent words are grouped into a segment. and the segments are also classified using the word class information. The same amount of information is inserted into each of the segment classes. The signal is encoded by modifying some inter-word spaces statistics of segment classes. Subjective comparisons with conventional word-shift algorithms are presented under several criteria.

• The Journal of the Acoustical Society of Korea
• /
• v.8 no.4
• /
• pp.60-64
• /
• 1989

In this paper, an local path constraint Is proposed in order to increase the speech recognition rate. An input speech signal is analyzed by autocorrelation and LPC coefficient as parameters. The local path constraint of the proposed type was compared with the conventional five types. The speechs used in this search are the subway stops, and the 130 words pronounced 10 times for the different 13 words consisting of 11 characters of syllable by 2 male and 1 female are tested. As a result, we proved that this proposed type is the most optimal type and the recognition rate of $94.6\%$ is obtained .

• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.801-804
• /
• 2001

This article is concerned with automatic segmentation of two adjacent vowels for speech signals. All kinds of transition case of adjacent vowels can be characterized by spectrogram. Firstly the voiced-speech is extracted by the histogram analysis of vowel indicator which consists of wavelet low pass components. Secondly given phonetic transcription and transition pattern spectrogram, the voiced-speech portion which has consecutive vowels automatically segmented by the template matching. The cross-correlation function is adapted as a template matching method and the modified correlation coefficient is calculated for all frames. The largest value on the modified correlation coefficient series indicates the boundary of two consecutive vowel sounds. The experiment is performed for 154 vowel transition sets. The 154 spectrogram templates are gathered from 154 words(PRW Speech DB) and the 161 test words(PBW Speech DB) which are uttered by 5 speakers were tested. The experimental result shows the validity of the method.