• Journal of Biomedical Engineering Research
• /
• v.44 no.2
• /
• pp.147-156
• /
• 2023

Efforts to employ smart home sensors to monitor the indoor activities of elderly single residents have been made to assess the feasibility of a safe and healthy lifestyle. However, the bathroom remains an area of blind spot. In this study, we have developed and evaluated a new edge computer device that can automatically detect water usage activities in the bathroom and record the activity log on a cloud server. Three kinds of sound as flushing, showering, and washing using wash basin generated during water usage were recorded and cut into 1-second scenes. These sound clips were then converted into a 2-dimensional image using MEL-spectrogram. Sound data augmentation techniques were adopted to obtain better learning effect from smaller number of data sets. These techniques, some of which are applied in time domain and others in frequency domain, increased the number of training data set by 30 times. A deep learning model, called CRNN, combining Convolutional Neural Network and Recurrent Neural Network was employed. The edge device was implemented using Raspberry Pi 4 and was equipped with a condenser microphone and amplifier to run the pre-trained model in real-time. The detected activities were recorded as text-based activity logs on a Firebase server. Performance was evaluated in two bathrooms for the three water usage activities, resulting in an accuracy of 96.1% and 88.2%, and F1 Score of 96.1% and 87.8%, respectively. Most of the classification errors were observed in the water sound from washing. In conclusion, this system demonstrates the potential for use in recording the activities as a lifelog of elderly single residents to a cloud server over the long-term.

• Journal of the Institute of Convergence Signal Processing
• /
• v.16 no.4
• /
• pp.134-138
• /
• 2015

Heart sound is used for a basic clinical examination to check for abnormalities in the lungs and heart that can be heard with a stethoscope or phonocardiography. In this paper, we try to find an easier and non-invasive method to diagnose heart diseases using neural network classifier. The classifier has been developed for one normal heart sound and five murmurs by using Shannon entropy and conjugate scaled back propagation algorithm. The experimental results showed that the classification is possible with 1.63185e-6 of classification error.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2006.11a
• /
• pp.671-674
• /
• 2006

Heavy-weight impact sounds generated by impact ball were classified according to the frequency characteristics on the equal loudness contours. Sound quality metrics such as Zwicker's loudness, sharpness, roughness of each classified impact sound were also measured. Loudness spectrum has been regarded as an indication of the characteristics difference of each classified impact sound. The adjectives in Korean expressing the sound quality characteristics of floor impact sounds were also investigated by adoptability and similarity tests. The group of the adjectives was used to evaluate the sound quality of floor impact sound by semantic differential test method.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.16 no.5 s.110
• /
• pp.537-543
• /
• 2006

The impact sound of the golf club is one of the major factors to purchase it. Sound quality metrics has been mostly developed for harmonic sounds. Sound quality evaluation techniques for the impact sound have been contrived in this study. Jury test, one of the typical subjective evaluation scheme, is carried out for evaluating the sound quality of 13 different golf drivers, which classification results are assumed to be correct answers. Conventional objective evaluation methods such as Zwicker loudness sensory pleasantness are calculated. Wavelet analysis and instantaneous loudness are applied in order to evaluate the sound quality of transient sounds, which scheme shows better correlation with the results from jury test.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2005.05a
• /
• pp.595-598
• /
• 2005

The impact sound of the golf club is one of the major factors to purchase it. Sound qualify evaluation techniques are mostly developed for harmonic sounds. Sound quality metrics for the impact sound is proposed in this study. Jury test, one of the typical subjective evaluation scheme, is carried out for evaluating the sound quality of 11 different golf drivers. Above subjective classification results are assumed to be the right answers. Conventional objective evaluation methods such as Zwicker loudness sensory pleasantness are calculated. Wavelet analysis and instantaneous loudness are applied in order to evaluate the transient sounds, which shows better correlation with the results from those by jury test.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2006.05a
• /
• pp.722-725
• /
• 2006

Studies on the ways to construct agreeable sound-amenity have been processed in Korea recently and Virtual Acoustics Field Simulation System (VAFSS) which is an active acoustics reproducing system has been made as a technique to realize the results of the study. This system catches the changes of surroundings and produce sounds which go well with the mood of the space. The fact that a man thinks a sound goes well with factors of the environment should be an individual evaluation. Thus, the standards to classify factors influencing the preference of the sound, which can be judged by the environment, are needed. This study suggests the standards of factors to provide agreeable sound for people according to changes of the time and other elements. Among the factors influencing environment, the temperature, the humidity and the wind were suggested as standards of discomfort Index and wind chin temperature. Besides, only the intensity of illumination has been chosen to estimate the intensity of radiation as a part of factors of the whether.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.2
• /
• pp.143-148
• /
• 2016

Deep neural network that effectively capture the characteristics of data has been widely used in various applications. However, the amount of sound database is often insufficient for learning the deep neural network properly, so resulting in overfitting problems. In this paper, we propose a transfer learning framework that can effectively train the deep neural network even with insufficient sound event data by employing rich speech or music data. A series of experimental results verify that proposed method performs significantly better than the baseline deep neural network that was trained only with small sound event data.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.65 no.7
• /
• pp.1263-1269
• /
• 2016

Hearing impaired persons are exposed to the danger since they can't be aware of many dangerous situations like fire alarms, car hones and so on. Therefore they need haptic or visual informations when they meet dangerous situations. In this paper, we design a dangerous sound detection engine for hearing impaired. We consider four dangerous indoor situations such as a boiled sound of kettle, a fire alarm, a door bell and a phone ringing. For outdoor, two dangerous situations such as a car horn and a siren of emergency vehicle are considered. For a test, 6 data sets are collected from those six situations. we extract LPC, LPCC and MFCC as feature vectors from the collected data and compare the vectors for feasibility. Finally we design a matching engine using an artificial neural network and perform classification tests. We perform classification tests for 3 times considering the use outdoors and indoors. The test result shows the feasibility for the dangerous sound detection.

• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.2
• /
• pp.15-25
• /
• 2024

For effective analysis of animal ecosystems, technology that can automatically identify the current status of animal habitats is crucial. Specifically, animal sound classification, which identifies species based on their sounds, is gaining great attention where video-based discrimination is impractical. Traditional studies have relied on a single deep learning model to classify animal sounds. However, sounds collected in outdoor settings often include substantial background noise, complicating the task for a single model. In addition, data imbalance among species may lead to biased model training. To address these challenges, in this paper, we propose an animal sound classification scheme that combines predictions from multiple models using Focal Loss, which adjusts penalties based on class data volume. Experiments on public datasets have demonstrated that our scheme can improve recall by up to 22.6% compared to an average of single models.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.3
• /
• pp.144-150
• /
• 2006

Clinicians usually use stethoscopic auscultation for the diagnosis of heart diseases. However, the heart sound signal has varying characteristics due to the noise and/or the conditions of the patients. Also, it is not easy for junior clinicians to find the acoustical differences between different kinds or heart sound signals. which may result in errors in the diagnosis. Thus it will be quite useful for the clinicians to make use of an automatic classification system using signal processing techniques. In this paper, we propose to use hidden Markov models in stead of artificial neural networks which have been conventionally used for the automatic classification of heart sounds. In the experiments classifying heart sound signals. we could see that the proposed methods were quite successful in the classification accuracy.