• The Journal of the Acoustical Society of Korea
• /
• v.35 no.4
• /
• pp.303-309
• /
• 2016

This paper proposes a novel approach to the problem of audio signal clustering using a stacked autoencoder. The proposed stacked autoencoder learns an efficient representation for the input signal, enables clustering constituent signals with similar characteristics, and therefore the original sources can be separated based on the clustering results. STFT (Short-Time Fourier Transform) is performed to extract time-frequency spectrum, and rectangular windows at all the possible locations are used as input values to the autoencoder. The outputs at the middle, encoding layer, are used to cluster the rectangular windows and the original sources are separated by the Wiener filters derived from the clustering results. Source separation experiments were carried out in comparison to the conventional NMF (Non-negative Matrix Factorization), and the estimated sources by the proposed method well represent the characteristics of the orignal sources as shown in the time-frequency representation.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.15 no.8
• /
• pp.5280-5287
• /
• 2014

This paper proposes a method in which classes are grouped as a hierarchical tree structure for the effective classification of the Raman spectra. As experimental data, the Raman spectra of 28 chemical compounds were obtained, and pre-treated with noise removal and normalization. The spectra that induced a classification error were grouped into the same class and the hierarchical structure class was composed. Each high and low class was classified using a PCA-MAP method. According to the experimental results, the classification of 100% was achieved with 2.7 features on average when the proposed method was applied. Considering that the same classification rates were achieved with 6 features using the conventional method, the proposed method was found to be much better than the conventional one in terms of the total computational complexity and practical application.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.10
• /
• pp.1737-1743
• /
• 2008

A weather radar extracts the weather information from the return echoes which consist of scattered electromagnetic wave signals from rain, cloud and dust particles, etc. The characteristics of Doppler weather signal and ground clutter should be analyzed to extract the accurate weather information. However, the conventional symmetric weather Doppler model is somewhat inappropriate in representing various weather situations. Therefore, the improved model is suggested to describe the skewness in the Doppler spectrum model. Using the suggested model, many various weather signals can be simulated efficiently in time and spectral domain according to weather situations, operation environment and system characteristics. This simulation method may be very helpful in verifying the accuracy of the weather information extraction algorithms and developing the new system for further performance improvement.

• Proceedings of the Korea Concrete Institute Conference
• /
• 2010.05a
• /
• pp.113-114
• /
• 2010

In this study, the earthquake damage evaluation of a R/C frame building is carried out based on the method proposed in Part I. Using the proposed method, the earthquake damage of building system based on non-linear required strength spectrum can be effectively evaluated without using the detailed seismic evaluation methods, including non-linear dynamic analyses, capacity spectrum method, etc.

• Proceedings of the KIEE Conference
• /
• 2003.07d
• /
• pp.2031-2033
• /
• 2003

본 논문에서는 25%, 50%, 75%, 100% 정격 부하 아래에서 b유도전동기의 회전자 고장을 검출하기 위한 효과적인 FFT 기반 알고리즘을 제안하였다. 제안한 방법은 고정자 전류 스펙트럼 성분 중에서 회전자 고장에 큰 영향을 주는 주파수 성분에서 미리 결정한 기준벡터와 특정벡터 사이의 평균 절대치 차이(Mean Absolute Difference)를 이용하였다. 기준벡터는 정상 상태의 고정자 전류 스펙트럼 성분 중에서 기본 주파수 상, 하의 두개의 측파대 주변의 좁은 영역에서 추출하였고 특징벡터는 정상상태와 회전자 바 고장상태의 고정자 전류 스펙트럼 성분 중에서 또한 기준벡터와 동일한 영역에서 추출하였다. 부하실험을 통하여 제안한 알고리즘의 적용 결과는 각각의 정격 부하에서 유도전동기의 회전자 바 고장을 효과적으로 검출할 수 있음을 입증하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.7
• /
• pp.1370-1376
• /
• 2012

This paper proposes a gender recognition algorithm which classifies a male or female speaker. In this paper, characteristic vectors for the male and female speaker are analyzed, and recognition experiments for the proposed gender recognition by a neural network are performed using these characteristic vectors for the male and female. Input characteristic vectors of the proposed neural network are 10 LPC (Linear Predictive Coding) cepstrum coefficients, 12 LPC cepstrum coefficients, 12 FFT (Fast Fourier Transform) cepstrum coefficients and 1 RMS (Root Mean Square), and 12 LPC cepstrum coefficients and 8 FFT spectrum. The proposed neural network trained by 20-20-2 network are especially used in this experiment, using 12 LPC cepstrum coefficients and 8 FFT spectrum. From the experiment results, the average recognition rates obtained by the gender recognition algorithm is 99.8% for the male speaker and 96.5% for the female speaker.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.19 no.3
• /
• pp.297-303
• /
• 2009

In this paper, we propose the biomedical spectral pattern classification techniques by the fusion scheme based on the SpPCA and MLP in extended feature space. A conventional PCA technique for the dimension reduction has the problem that it can't find an optimal transformation matrix if the property of input data is nonlinear. To overcome this drawback, we extract features by the SpPCA technique in extended space which use the local patterns rather than whole patterns. In the classification step, individual classifier based on MLP calculates the similarity of each class for local features. Finally, biomedical spectral patterns is classified by the fusion scheme to effectively combine the individual information. As the simulation results to verify the effectiveness, the proposed method showed more improved classification results than conventional methods.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.6
• /
• pp.94-99
• /
• 1997

This paper describes a technique of probabilistic spectral subtraction which uses the knowledge of both noise and speech so as to reduce automatic speech recognition errors in noisy environments. Spectral subtraction method estimates a noise prototype in non-speech intervals and the spectrum of clean speech is obtained from the spectrum of noisy speech by subtracting this noise prototype. Thus noise can not be suppressed effectively using a single noise prototype in case the characteristics of the noise prototype are different from those of the noise contained in input noisy speech. To modify such a drawback, multiple noise prototypes are used in probabilistic subtraction method. In this paper, the probabilistic characteristics of noise and the knowledge of speech which is embedded in hidden Markov models trained in clean environments are used to suppress noise. Futhermore, dynamic feature parameters are considered as well as static feature parameters for effective noise suppression. The proposed method reduced error rates in the recognition of 50 Korean words. The recognition rate was 86.25% with the probabilistic subtraction, 72.75% without any noise suppression method and 80.25% with spectral subtraction at SNR(Signal-to-Noise Ratio) 10 dB.

• Journal of the Institute of Convergence Signal Processing
• /
• v.1 no.1
• /
• pp.15-22
• /
• 2000

The analysis of power spectrum based on linear AR model is applied widely to quantize the response of autonomic nerve noninvasively, In this paper, we estimate the power spectrum density for heartrate variability of the electrocadiogram and pulse wave for short term data(less than two minute), The time series of heart rate variability is obtained from the time interval(RRI, PPI) between the feature point of the electrocadiogram and pulse wave for normal person, The generated time series reconstructed into new time series through polynomial interpolation to apply to the AR mode. The power spectrum density for AR model is calculated by Burg algorithm, After applying AR model, the power spectrum density for heart rate variability of the electrocadiogram and the pulse wave is shown smooth spectrum power at the region of low frequence and high frequence, and that the power spectrum density of electrocadiogram and pulse wave has similar form for same subject.

• Korean Journal of Soil Science and Fertilizer
• /
• v.42 no.6
• /
• pp.522-528
• /
• 2009

This study focused on establishing prediction models using visible-near infrared spectrum to simultaneously detect multiple components of soils and enhancing the performance quality by suitably transformed input spectra and classification of soil spectral types for prediction model input. The continuum-removed spectra showed significant result for all cases in terms of soil properties and classified or bulk predictions. The prediction model using classified soil spectra at an absorption peak area around 500nm and 950nm efficiently indicating soil color showed slightly better performance. Especially, Ca and CEC were well estimated by the classified prediction model at $R^{2}$ > 0.8. For organic carbon, both classified and bulk prediction model had a good performance with $R^{2}$ > 0.8 and RPD> 2. This prediction model may be applied in global soil mapping, soil classification, and remote sensing data analysis.