Full waveform inversion using a decomposed single frequency component from a spectrogram

Ha, Jiho,Kim, Seongpil,Koo, Namhyung,Kim, Young-Ju,Woo, Nam-Sub,Han, Sang-Mok,Chung, Wookeen,Shin, Sungryul,Shin, Changsoo,Lee, Jaejoon Elsevier 2018 Journal of applied geophysics Vol.153 No.-

<P><B>Abstract</B></P> <P>Although many full waveform inversion methods have been developed to construct velocity models of subsurface, various approaches have been presented to obtain an inversion result with long-wavelength features even though seismic data lacking low-frequency components were used. In this study, a new full waveform inversion algorithm was proposed to recover a long-wavelength velocity model that reflects the inherent characteristics of each frequency component of seismic data using a single-frequency component decomposed from the spectrogram. We utilized the wavelet transform method to obtain the spectrogram, and the decomposed signal from the spectrogram was used as transformed data. The Gauss–Newton method with the diagonal elements of an approximate Hessian matrix was used to update the model parameters at each iteration. Based on the results of time–frequency analysis in the spectrogram, numerical tests with some decomposed frequency components were performed using a modified SEG/EAGE salt dome (A–A′) line to demonstrate the feasibility of the proposed inversion algorithm. This demonstrated that a reasonable inverted velocity model with long-wavelength structures can be obtained using a single frequency component. It was also confirmed that when strong noise occurs in part of the frequency band, it is feasible to obtain a long-wavelength velocity model from the noise data with a frequency component that is less affected by the noise. Finally, it was confirmed that the results obtained from the spectrogram inversion can be used as an initial velocity model in conventional inversion methods.</P> <P><B>Highlights</B></P> <P> <UL> <LI> We proposed the full waveform inversion method using spectral decomposition method. </LI> <LI> The decomposed signal from spectrogram has the reproduced low-frequency components. </LI> <LI> The results show that the inverted velocity models have long-wavelength features. </LI> <LI> The inverted results have different shapes varying with decomposed components. </LI> </UL> </P>

Power Signal Classification with Combinational Spectrogram-based CNN for Embedded System Health Management

Heoncheol Lee,Yongsung Kwon,Kipyo Kim 제어로봇시스템학회 2018 제어로봇시스템학회 국제학술대회 논문집 Vol.2018 No.10

This paper addresses the problem of the embedded system health management for high-speed flight systems. Especially, we focus the variation of power signals used in embedded systems because the electrical degeneration is strongly related to the power levels and frequencies. If the power signals can be classified into normal status and abnormal status, the sudden electrical degeneration of embedded systems can be successfully detected. The conventional threshold-based classification which has been used in aerospace and defense fields cannot find out the hidden anomaly within the thresholds. This paper proposes an accurate power signal classification method using combinational spectrogram-based convolutional neural networks (CNN). The power signals are combined with eigenvalues and converted to spectrogram which can analyze them on time and frequency domain simultaneously. Then, the CNN for power signal classification is trained and validated using the combinational spectrograms. Inference results showed that the proposed method can accurately classify the power signals into normal status and abnormal status.

Spectrographic analysis of normal and adventitious breath sounds in children

( Sung Eun Kim ),( Jong-seo Yoon ) 대한결핵 및 호흡기학회 2019 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.127 No.-

Objective: Diagnosing pulmonary diseases with auscultation of breath sounds in children is often difficult because of its high dependence on the clinician’s experience of special circumstances involving children. The purpose of this study was to analyze the spectrograms of normal and adventitious breath sounds to reproducibly find the appropriate character of breath sounds through a mechanical analysis. Method: Breath sounds of 90 children and adolescents aged 0-14 years (median: 4.5 years) were analyzed. They were recorded with an electronic stethoscope (JABES®). Spectrograms of all breath sounds were drawn, and the characteristics of the spectrograms were studied. Result: A total of 173 records of breath sounds of 90 subjects were categorized into normal, wheezing, rhonchi, wheezing combined with rhonchi, crackling, groaning, squawking, and snoring sounds, including 116, 12, 17, 9, 12, 5, 1, and 1 records, respectively. The spectrogram of normal breath sounds showed a diffuse frequency distribution forming a typical dome shape. For wheezing, rhonchi, and their combination, one or more specific descent lines of certain frequencies were observed during expiration. The lines were not horizontal, unlike previously reported. For crackles, the amplitude was high at certain frequencies, forming a scattered distribution of dots. For groaning sound, multiple polymorphic lines with a high amplitude were tangled during expiration. For squawking sound, various forms of multiple lines with different frequencies were seen at the end of inspi ration. For snoring sound, the frequency was widely dispersed with a high amplitude during inspiration and showed multiple descent lines during expiration. Conclusion: Typical characteristics of normal and adventitious breath sounds were determined from the corresponding spectrograms. Spectrography of more samples would improve our understanding of the characteristics of different breath sounds that are not wellknown to date.

R-Net: An LSTM-Assisted Deep Learning Framework for the Classification of Bird Species Based on Sound-Spectrogram in Rambutan Agriculture Field

Theresa Jose,J Albert Mayan 한국인터넷정보학회 2025 KSII Transactions on Internet and Information Syst Vol.19 No.11

Accurate bird species classification is as crucial to monitoring biodiversity and ecological studies as it is in agricultural settings, where birds have both positive and negative impacts on crop production. The proposed research presents a Long Short-Term Memory (LSTM) assisted deep learning framework, RambutanNet (R-Net), designed to classify bird species from their sound spectrograms. R-Net utilises the capabilities of LSTM, integrated with neural networks, to efficiently extract temporal and spatial features from spectrograms of calls. This hybrid model effectively captures both local frequency patterns and long-range dependencies of the audio signals, which is crucial for differentiating between species with similar acoustic signatures. The spectrogram dataset is created by recording sounds from birds using high-fidelity audio equipment and sourced through various repositories. The recordings were pre-processed to remove noise and irrelevant environmental sounds, then converted into mel-spectrograms. The spectral content is fed into the R-Net framework to perform feature extraction and LSTM layers that model the temporal relationships in the audio sequences. The R-Net model was developed and deployed in a Raspberry Pi 5 single-board computer, and its validation and demonstration of performance in multiple fields were conducted. The model achieved an accuracy of 97.86%, along with precision and recall of 98.03% and 96.15%, respectively, showing the extensibility of the model that distinguishes bird species even in extreme natural environments. Another measure is the F1 score, which balances precision with recall and proved to be 98.34%. The substantial AUROC value of 0.95 also points to the model's accurate ability to differentiate among bird species according to their acoustic signatures.

A Spectrogram Based Local Fluctuation Feature for Fault Diagnosis with Application to Rotating Machines

Jiang Qinyu,Chang Faliang,Liu Chunsheng 대한전기학회 2021 Journal of Electrical Engineering & Technology Vol.16 No.4

Rotating machines are one of the most common equipment in modern industry, the health condition of the equipment is closely linked to safety of workers and production eff ectiveness. Thus accurate and robust fault diagnostic approaches are vital to safety production. In practice, diagnostic accuracy is seriously aff ected by noises, especially in low signal-to-noise (SNR) ratio conditions, and the quality of fault features is positively link to the diagnosing accuracy. In consideration of distinguishable feature expression can improve diagnosing result and robust to wider range of experimental conditions, this paper presents a novel spectrogram based local fl uctuation feature (SLFF) for low SNR conditions. Firstly, signals are transformed into spectrograms. Then a feature extracting window bank is established on spectrograms for SLFF. At last, a support vector machine (SVM) is applied as a fault classifi er for evaluating the proposed feature. The proposed SLFF represents the basic spectral shape and variation which leads to robust and well distinguishable feature expression, the feature reveals the diff erences of spectral local variation trends between normal and fault types that can improve the discrimination under the infl uence of strong noises. The eff ectiveness of the proposed method has been proved experimentally in this paper.

Optimized deep neural network models for blood pressure classification using Fourier analysis‑based time–frequency spectrogram of photoplethysmography signal

Pankaj,Ashish Kumar,Manjeet Kumar,Rama Komaragiri 대한의용생체공학회 2023 Biomedical Engineering Letters (BMEL) Vol.13 No.4

Appropriate blood pressure (BP) management through continuous monitoring and rapid diagnosis helps to take preventivecare against cardiovascular diseases (CVD). As hypertension is one of the leading causes of CVDs, keeping hypertensionunder control by a timely screening of subjects becomes lifesaving. This work proposes estimating BP from motion artifactaffectedphotoplethysmography signals (PPG) by applying signal processing techniques in realtime. This paper proposes adeep neural network-based methodology to accurately classify PPG signals using a Fourier theory-based time–frequency(TF) spectrogram. This work uses the Fourier decomposition method (FDM) to transform a PPG signal into a TF spectrogram. In the proposed work, the last three layers of the pre-trained deep neural network, namely, GoogleNet, DenseNet, andAlexNet, are modified and then used to classify the PPG signal into normotension, pre-hypertension, and hypertension. Theproposed framework is trained and tested using the MIMIC-III and PPG–BP databases using five-fold training and testing. Out of the three deep neural networks, the proposed framework with the DenseNet-201 network performs best, with a testaccuracy of 96.5%. The proposed work uses FDM to compute the TF spectrogram to accurately separate the motion artifactsand noise components from a noise-corrupted PPG signal. Capturing more frequency components that contain moreinformation from PPG signals makes the deep neural networks extract better and more meaningful features. Thus, training adeep neural network model with clean PPG signal features improves the generalized capability of a BP classification modelwhen tested in realtime.

RTO Rotary Motor의 고장 예측을 위한 스펙트로그램 기반 학습 알고리즘 구현 - 진동 분석을 통한 알고리즘 연구 -

박훈민,임태영,김언규,민흥기,윤달환,김동원,김지민,김성준 사단법인 한국안전문화학회 2025 안전문화연구 Vol.- No.44

The purpose of this study is to develop an AI-based diagnostic system capable of accurately predicting failures of the rotary motor used in regenerative thermal oxidizer (RTO) equipment. The system utilizes vibration data, which is critical for machinery condition monitoring. To achieve this, vibration and phase current data of the motor were collected using an RC low-pass filter with a cutoff frequency of 1 kHz, and the data were sampled at a rate of 1 kHz. The collected data were transformed into spectrograms via Fast Fourier Transform (FFT) and used as input for the model. A deep learning model based on a ResNet architecture was designed and trained to predict motor conditions using the spectrograms. The model’s performance was evaluated using key metrics such as accuracy and loss. Experimental results demonstrate that the proposed algorithm effectively analyzes vibration signals and accurately predicts the rotary motor’s condition. This outcome suggests the potential application of AI-based diagnostic systems in real-time monitoring and preventive maintenance of RTO equipment. In particular, by enabling early fault detection of rotary motors—which control the inflow and outflow of gases in environments exceeding 800°C—the proposed system significantly improves safety compared to conventional field operations.

Exploring the Relation Between EMG Pattern Recognition and Sampling Rate Using Spectrogram

Jingwei Too,Abdul Rahim Abdullah,Norhashimah Mohd Saad,Nursabillilah Mohd Ali,Tengku Nor Shuhada Tengku Zawawi 대한전기학회 2019 Journal of Electrical Engineering & Technology Vol.14 No.2

The application of electromyography (EMG) has shown great success in rehabilitation engineering. With the existing multiple- channel EMG recording system, the detection and classifi cation of EMG pattern have become viable. The purpose of this study is to investigate the relation between sampling rate and EMG pattern recognition by using spectrogram. The features are extracted from spectrogram coeffi cients and the principal component analysis is applied for dimensionality reduction. In addition, the optimal Hanning window size is identifi ed and selected before performance evaluation. For noise evaluation, the additive white Gaussian noise (AGWN) is added to the EMG signal at 30, 25, 20 dB SNR. The results illustrated that the 512 Hz sampling rate can maintain a small decrement of 0.76% accuracy compared to 1024 Hz. However, when the AGWN is added, the 256 and 512 Hz sampling rates showed a greater reduction in overall classifi cation performance. For a lower SNR, the gaps in classifi cation accuracy between 1024 Hz, 512 Hz and 256 Hz sampling rates are obviously presented. It signifi es that reducing the sampling rate lower than 1024 Hz might not be a good choice since the noise and artifact have to be taken into consideration in a real system.

텍스트와 음성의 앙상블을 통한 다중 감정인식 모델

신주현,임명진,이명호 (사)한국스마트미디어학회 2022 스마트미디어저널 Vol.11 No.8

Due to COVID-19, the importance of non-face-to-face counseling is increasing as the face-to-face counseling method has progressed to non-face-to-face counseling. The advantage of non-face-to-face counseling is that it can be consulted online anytime, anywhere and is safe from COVID-19. However, it is difficult to understand the client's mind because it is difficult to communicate with non-verbal expressions. Therefore, it is important to recognize emotions by accurately analyzing text and voice in order to understand the client's mind well during non-face-to-face counseling. Therefore, in this paper, text data is vectorized using FastText after separating consonants, and voice data is vectorized by extracting features using Log Mel Spectrogram and MFCC respectively. We propose a multi-emotion recognition model that recognizes five emotions using vectorized data using an LSTM model. Multi-emotion recognition is calculated using RMSE. As a result of the experiment, the RMSE of the proposed model was 0.2174, which was the lowest error compared to the model using text and voice data, respectively. COVID-19로 인해 대면으로 이루어지던 상담 방식이 비대면으로 진행되면서 비대면 상담의 중요성이 높아지고 있다. 비대면 상담은 온라인으로 언제 어디서든 상담할 수 있고, COVID-19에 안전하다는 장점이 있다. 그러나 비언어적 표현의 소통이 어려워 내담자의 마음을 이해하기 어렵다. 이에 비대면 상담 시 내담자의 마음을 잘 알기 위해서는 텍스트와 음성을 정확하게 분석하여 감정을 인식하는 것이 중요하다. 따라서 본 논문에서는 텍스트 데이터는 자음을 분리한 후 FastText를 사용하여 벡터화하고, 음성 데이터는 Log Mel Spectrogram과 MFCC를 사용하여 각각 특징을 추출하여 벡터화한다. 벡터화된 데이터를 LSTM 모델을 활용하여 5가지 감정을 인식하는 다중 감정인식 모델을 제안한다. 다중 감정인식은 RMSE을 활용하여 계산한다. 실험 결과 텍스트와 음성 데이터를 각각 사용한 모델보다 제안한 모델의 RMSE가 0.2174로 가장 낮은 오차를 확인하였다.

머신러닝 기반의 선박 방사 소음 스펙트로그램 분리 및 로이드미러 패턴 추출 기법

김보규,김도훈,신승국,최지웅,한동균,김대경 한국군사과학기술학회 2026 한국군사과학기술학회지 Vol.29 No.2

- This study presents an integrated machine learning framework for separating narrowband components and extracting Lloyd's mirror interference patterns from ship-radiated noise(SRN) spectrograms. The proposed methodology employs Independent Vector Analysis to separate narrowband spectral components from multi-hydrophone acoustic signals, subsequently applying DBSCAN and RANSAC algorithms for robust identification of parabolic Lloyd's mirror patterns in residual spectrograms. Experimental validation utilizing SRN data acquired during the SAVEX-15 sea trials demonstrates effective narrowband component separation, as verified through DEMON and LOFAR analyses, alongside accurate pattern extraction capabilities. The unsupervised framework exhibits enhanced reliability under adverse noise conditions and enables precise closest point of approach(CPA) estimation. The developed methodology offers an automated and robust solution for SRN analysis, significantly improving acoustic signal interpretation and target identification capabilities in maritime defense applications.