• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.682-685
• /
• 2004

This paper investigated a method for classifying emotional states by using pulse wave signal. It focused on finding effective features for emotional state classification. The emptional states considered here consisted of interest and neutral. Classification experiments utilized 65 and 60 samples of interest and neutral states respectively. We have investigated 19 features derived from pulse wave signals by using both time domain and frequency domain analysis methods with 2 classifiers of minimum distance (normalized Euclidean distanece) and ${\kappa}$-Nearest Neighbour. The Leave-one-out cross validation was used as an evaluation mehtod. Based on experimental results, the most efficient features were a combination of 4 features consisting of (i) the mean of the first differences of the smoothed pulse rate time series signal, (ii) the mean of absolute values of the second differences of thel normalized interbeat intervals, (iii) the root mean square successive difference, and (iv) the power in high frequency range in normalized unit, which provided 80.8% average accuracy with ${\kappa}$-Nearest Neighbour classifier.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.11
• /
• pp.53-61
• /
• 1996

In this paper, efficient techniques for solder joint inspection have been described. Using three layers of ring shaped LED's with different illumination angles, three frames of images are sequentially obtained. From these images the regions of interest (soldered regions) are segmented, and their characteristic features including the average gray level and the percentage of highlights - refereed to as 2D features - are extracted. Based on the backpropagation algorithm of neural networks, each solder joint is classified intor one of the pre-defined types. If the output value is not in the confidence interval, the distribution of tilt angles-referred to as 3D features - is claculated, and the solder joint is classified based on the bayes classfier. The second classifier requires more computation while providing more information and better performance. The proposed inspection system has been implemented and tested with various types of solder joints in SMDs. The experimental results have verified the validity of this scheme in terms of speed and recognition rate.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1993.06a
• /
• pp.125-127
• /
• 1993

오디오 혹은 비디오화의, 방송 고품질전화 등의 음성신호의 전송을 위해 마련된 CCITT Recommendation G.722에 의거 Codec을 구성하고 이를 통과한 연속음성을 CMU의 불특정 화자 연속음성인식 시스템인 SPHINX에 입력하여 인식률을 조사 한 후 CODING전의 인식결과와 비교하였다. 이때 CODEC은 크게 네 부분(Trans Quarature Mirror Filter, Encoder, Decoder, Receive QMF)으로 구성하고 입력음성 데이터는 150화자에 의한 1018문장을 훈련용으로, 140문장을 테스트용으로 하였을 때의 단어 인식률을 인식률로 하였다. 또 이때 특징벡터로는 12차 Melcepstrum 계수를 사용하였다. 인식결과 코딩전(close talk Mic를 이용하여 직접입력)의 단어 인식률이 86.7%인데 비해 코딩후의 인식률은 85.6%로 나타나 약 1%의 인식률 저하를 가져와 코딩으로 인한 Error에 비해 비교적 양호한 결과를 얻을 수 있었다. 인식률 저하의 원인으로서는 코딩시의 BER(Bit Error Rate)에 의한 것으로 생각된다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1985.10a
• /
• pp.75-78
• /
• 1985

본 연구는 시간 영역의 parament를 이용하여 한국어 숫자음(영, 일, 이, 삼, 사, 오, 육, 칠, 팔, 구)을 인식했다. 입력 음성 신호 X(n)의 Beginning Point와 Ending point를 ZCR(Zero-crossing Rate), Magnitude, Energy, Autocorrelation을 이용 Beginning point와 Ending point를 구하고 자음부의 인식은 위 계수들을 이용하여 행했다. 또, 유성음 부분에서는 PARCOR(Partial Autocorrelation), LPC(Linear Predictive Coding)를 이용 모음부와 유성자음을 인식하여 모음을 6개 부류(ㅏ, ㅑ, ㅗ, ㅜ, ㅠ, ㅣ)로 구분 인식했다. 이 방법에 의하면 입력 음성 신호 X(n)의 B.P(Beginning Point)와 E.P(Ending Point)를 쉽게 추출 가능하며 또한 각 Parameter를 이용하여 94.4%의 인식율을 얻었다.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.49 no.12
• /
• pp.651-658
• /
• 2000

This paper presents a new method to extract traffic information such as number of passing vehicles and average speed by a pair of stereo road images. The whole process consists of the extraction of vehicles and the tracking of the extracted vehicles. For the extraction of vehicles, the outline of each vehicle is obtained by using binary region growing technique applied to disparity map based on multi-resolution stereo matching. The Kalman filter tracking algorithm is applied to the extracted vehicle outlines to determine the flow of vehicles. Experimental results show that the proposed method significantly improved recognition rate of vehicles over the conventional methods-frame difference and background elimination methods.

• Journal of Biomedical Engineering Research
• /
• v.28 no.2
• /
• pp.205-211
• /
• 2007

A novel envelope extraction algorithm for speech processor of cochlear implants, called adaptation algorithm, was developed which is based on a adaptation effect of the inner hair cell(IHC)/auditory nerve(AN) synapse. We achieved acoustic simulation and hearing experiments with 12 normal hearing persons to compare this adaptation algorithm with existent standard envelope extraction method. The results shows that speech processing strategy using adaptation algorithm showed significant improvements in speech recognition rate under most channel/noise condition, compared to conventional strategy We verified that the proposed adaptation algorithm may yield better speech perception under considerable amount of noise, compared to the conventional speech processing strategy.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.395-398
• /
• 1998

In this paper, a method for the improvement of learning speed and convergence rate was proposed applied it to physiological neural structure with the advantages of artificial neural networks and fuzzy theory to physiological neuron structure, To compare the proposed method with conventional the single layer perception algorithm, we applied these algorithms bit parity problem and pattern recognition containing noise. The simulation result indicated that our learning algorithm reduces the possibility of local minima more than the conventional single layer perception does. Furthermore we show that our learning algorithm guarantees the convergence.

• Speech Sciences
• /
• v.14 no.1
• /
• pp.163-174
• /
• 2007

All observation has uncertainty due to noise or channel characteristics. This uncertainty should be counted in the modeling of observation. In this paper we propose a modified optimization object function of a GMM training considering inexact observation. The object function is modified by introducing the concept of observation confidence as a weighting factor of probabilities. The optimization of the proposed criterion is solved using a common EM algorithm. To verify the proposed method we apply it to the speaker recognition domain. The experimental results of text-independent speaker identification with VidTimit DB show that the error rate is reduced from 14.8% to 11.7% by the modified GMM training.

• Speech Sciences
• /
• v.15 no.1
• /
• pp.75-83
• /
• 2008

The interior of an automobile is a very noisy environment with both stationary cruising noise and the reverberated music or speech coming out from the audio system. For robust speech recognition in a car environment, it is necessary to extract a driver's voice command well by removing those background noises. Since we can handle the music and speech signals from an audio system in a car, the reverberated music and speech sounds can be removed using an acoustic echo canceller. In this paper, we implement an acoustic echo canceller with robust double-talk detection algorithm using TMS-320C6711 DSK. First we developed the echo canceller on the PC for verifying the performance of echo cancellation, then implemented it on the TMS320C6711 DSK. For processing of one speech sample with 8kHz sampling rate and 256 filter taps of the echo canceller, the implemented system used only 0.035ms and achieved the ERLE of 20.73dB.

• Phonetics and Speech Sciences
• /
• v.5 no.1
• /
• pp.11-16
• /
• 2013

This paper proposes robust feature extraction for accurate voice activity detection (VAD). VAD is one of the principal modules for speech signal processing such as speech codec, speech enhancement, and speech recognition. Noisy environments contain nonstationary noises causing the accuracy of the VAD to drastically decline because the fluctuation of features in the noise intervals results in increased false alarm rates. In this paper, in order to improve the VAD performance, harmonic-weighted energy is proposed. This feature extraction method focuses on voiced speech intervals and weighted harmonic-to-noise ratios to determine the amount of the harmonicity to frame energy. For performance evaluation, the receiver operating characteristic curves and equal error rate are measured.