Milton Garces
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Milton Garces has a total of 70 co-authors. Here are the top co-authors by number of shared publications:
- Sarah K. Popenhagen
Infrasound Laboratory, University of Hawai’i at Mānoa, Honolulu, HI 96822, USA
7 shared publications - Samuel Kei Takazawa
Infrasound Laboratory, Hawaiʻi Institute of Geophysics and Planetology, School of Ocean and Earth Science and Technology, University of Hawaiʻi at Mānoa, Kailua-Kona, HI 96740, USA
6 shared publications - David Fee
Wilson Alaska Technical Center, Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska, U.S.A.
5 shared publications C. Hetzer
Hawaii Institute of Geophysics and Planetology, School of Ocean and Earth Science and Technology University of Hawaii at Manoa Honolulu Hawaii USA
4 shared publications- David L. Chichester
Idaho National Laboratory, Idaho Falls, USA
4 shared publications - Douglas P. Drob
U.S. Naval Research Laboratory Washington DC USA
4 shared publications
Latest Publications
- Open Access
Rocket Launch Detection with Smartphone Audio and Transfer Learning
Rocket launches generate infrasound signatures that have been detected at great distances. Due to the sparsity of the networks that have made these detections, however, most signals are detected tens of minutes to hours after the rocket launch. In this work, a method of near-real-time detection of rocket launches using data from a network of smartphones located 10–70 km from launch sites is presented. A machine learning model is trained and tested on the open-access Aggregated Smartphone Timeseries of Rocket-generated Acoustics (ASTRA), Smartphone High-explosive Audio Recordings Dataset (SHAReD), and ESC-50 datasets, resulting in a final accuracy of 97% and a false positive rate of <1%. The performance and behavior of the model are summarized, and its suitability for persistent monitoring applications is discussed.
Open AccessRocket Launch Detection with Smartphone Audio and Transfer Learning
Rocket launches generate infrasound signatures that have been detected at great distances. Due to the sparsity of the networks that have made these detections, however, most signals are detected tens of minutes to hours after the rocket launch. In this work, a method of near-real-time detection of rocket launches using data from a network of smartphones located 10–70 km from launch sites is presented. A machine learning model is trained and tested on the open-access Aggregated Smartphone Timeseries of Rocket-generated Acoustics (ASTRA), Smartphone High-explosive Audio Recordings Dataset (SHAReD), and ESC-50 datasets, resulting in a final accuracy of 97% and a false positive rate of <1%. The performance and behavior of the model are summarized, and its suitability for persistent monitoring applications is discussed.
- Open Access
Acoustic Rocket Signatures Collected by Smartphones
Rockets generate complex acoustic signatures that can be detected over a thousand kilometers from their source. While many far-field acoustic rocket signatures have been collected and released to the public, very few signatures collected at distances less than 100 km are available. This work presents a curated and annotated dataset of acoustic signatures of 243 rocket launches collected by a network of smartphones stationed at distances between 10 and 70 km from the launch sites, resulting in 1089 individual recordings. Due to the frequency dependence of atmospheric attenuation and the relatively short propagation distances, higher-frequency features not preserved in most publicly available data are observed. The signals are time-aligned to allow for different segments of the signal (ignition, launch, trajectory, chronology) to be more easily examined and compared. Initial analysis of the features of these rocket launch stages is performed, observed features are compared to those found in the existing literature, and comparisons between signals from launches of different rocket types are made. The dataset is annotated and made available to the public to aid future analysis of the characteristics and source mechanisms of rocket acoustics as well as applications such as rocket detection and classification models.
Open AccessAcoustic Rocket Signatures Collected by Smartphones
Rockets generate complex acoustic signatures that can be detected over a thousand kilometers from their source. While many far-field acoustic rocket signatures have been collected and released to the public, very few signatures collected at distances less than 100 km are available. This work presents a curated and annotated dataset of acoustic signatures of 243 rocket launches collected by a network of smartphones stationed at distances between 10 and 70 km from the launch sites, resulting in 1089 individual recordings. Due to the frequency dependence of atmospheric attenuation and the relatively short propagation distances, higher-frequency features not preserved in most publicly available data are observed. The signals are time-aligned to allow for different segments of the signal (ignition, launch, trajectory, chronology) to be more easily examined and compared. Initial analysis of the features of these rocket launch stages is performed, observed features are compared to those found in the existing literature, and comparisons between signals from launches of different rocket types are made. The dataset is annotated and made available to the public to aid future analysis of the characteristics and source mechanisms of rocket acoustics as well as applications such as rocket detection and classification models.
- Open Access
Explosion Detection Using Smartphones: Ensemble Learning with the Smartphone High-Explosive Audio Recordings Dataset and the ESC-50 Dataset
Explosion monitoring is performed by infrasound and seismoacoustic sensor networks that are distributed globally, regionally, and locally. However, these networks are unevenly and sparsely distributed, especially at the local scale, as maintaining and deploying networks is costly. With increasing interest in smaller-yield explosions, the need for more dense networks has increased. To address this issue, we propose using smartphone sensors for explosion detection as they are cost-effective and easy to deploy. Although there are studies using smartphone sensors for explosion detection, the field is still in its infancy and new technologies need to be developed. We applied a machine learning model for explosion detection using smartphone microphones. The data used were from the Smartphone High-explosive Audio Recordings Dataset (SHAReD), a collection of 326 waveforms from 70 high-explosive (HE) events recorded on smartphones, and the ESC-50 dataset, a benchmarking dataset commonly used for environmental sound classification. Two machine learning models were trained and combined into an ensemble model for explosion detection. The resulting ensemble model classified audio signals as either “explosion”, “ambient”, or “other” with true positive rates (recall) greater than 96% for all three categories.
Open AccessExplosion Detection Using Smartphones: Ensemble Learning with the Smartphone High-Explosive Audio Recordings Dataset and the ESC-50 Dataset
Explosion monitoring is performed by infrasound and seismoacoustic sensor networks that are distributed globally, regionally, and locally. However, these networks are unevenly and sparsely distributed, especially at the local scale, as maintaining and deploying networks is costly. With increasing interest in smaller-yield explosions, the need for more dense networks has increased. To address this issue, we propose using smartphone sensors for explosion detection as they are cost-effective and easy to deploy. Although there are studies using smartphone sensors for explosion detection, the field is still in its infancy and new technologies need to be developed. We applied a machine learning model for explosion detection using smartphone microphones. The data used were from the Smartphone High-explosive Audio Recordings Dataset (SHAReD), a collection of 326 waveforms from 70 high-explosive (HE) events recorded on smartphones, and the ESC-50 dataset, a benchmarking dataset commonly used for environmental sound classification. Two machine learning models were trained and combined into an ensemble model for explosion detection. The resulting ensemble model classified audio signals as either “explosion”, “ambient”, or “other” with true positive rates (recall) greater than 96% for all three categories.
- Open Access
- 1 September 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 156 (3) , 1509-1523
A comparison of smartphone and infrasound microphone data from a fuel air explosive and a high explosive
For prompt detection of large (>1 kt) above-ground explosions, infrasound microphone networks and arrays are deployed at surveyed locations across the world. Denser regional and local networks are deployed for smaller explosions, however, they are limited in number and are often deployed temporarily for experiments. With the expanded interest in smaller yield explosions targeted at vulnerable areas such as population centers and key infrastructures, the need for more dense microphone networks has increased. An “attritable” (affordable, reusable, and replaceable) and flexible alternative can be provided by smartphone networks. Explosion signals from a fuel air explosive (thermobaric bomb) and a high explosive with trinitrotoluene equivalent yields of 6.35 and 3.63 kg, respectively, were captured on both an infrasound microphone and a network of smartphones. The resulting waveforms were compared in time, frequency, and time-frequency domains. The acoustic waveforms collected on smartphones produced a filtered explosion pulse due to the smartphone's diminishing frequency response at infrasound frequencies (<20 Hz) and was found difficult to be used with explosion characterization methods utilizing waveform features (peak overpressure, impulse, etc.). However, the similarities in time frequency representations and additional sensor inputs are promising for other explosion signal identification and analysis. As an example, a method utilizing the relative acoustic amplitudes for source localization using the smartphone sensor network is presented.
- 1 September 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 156 (3) , 1509-1523
Open AccessA comparison of smartphone and infrasound microphone data from a fuel air explosive and a high explosive
For prompt detection of large (>1 kt) above-ground explosions, infrasound microphone networks and arrays are deployed at surveyed locations across the world. Denser regional and local networks are deployed for smaller explosions, however, they are limited in number and are often deployed temporarily for experiments. With the expanded interest in smaller yield explosions targeted at vulnerable areas such as population centers and key infrastructures, the need for more dense microphone networks has increased. An “attritable” (affordable, reusable, and replaceable) and flexible alternative can be provided by smartphone networks. Explosion signals from a fuel air explosive (thermobaric bomb) and a high explosive with trinitrotoluene equivalent yields of 6.35 and 3.63 kg, respectively, were captured on both an infrasound microphone and a network of smartphones. The resulting waveforms were compared in time, frequency, and time-frequency domains. The acoustic waveforms collected on smartphones produced a filtered explosion pulse due to the smartphone's diminishing frequency response at infrasound frequencies (<20 Hz) and was found difficult to be used with explosion characterization methods utilizing waveform features (peak overpressure, impulse, etc.). However, the similarities in time frequency representations and additional sensor inputs are promising for other explosion signal identification and analysis. As an example, a method utilizing the relative acoustic amplitudes for source localization using the smartphone sensor network is presented.
- Open Access
- 25 July 2024
- Published by MDPI AG in Preprints.org
Explosion Detection using Smartphones: Ensemble Learning with the Smartphone High-explosive Audio Recordings Dataset and the ESC-50 Dataset
Explosion monitoring is performed by infrasound and seismoacoustic sensor networks that are distributed globally, regionally, and locally. However, these networks are unevenly and sparsely distributed, especially in the local scale as maintaining and deploying networks is costly. With increasing interest in smaller yield explosions, the need for more dense networks has increased. To address this issue, we propose using smartphone sensors for explosion detection as they are cost-effective and easy to deploy. Although there are studies using smartphone sensors for explosion detection, the field is still in its infancy and new technologies need to be developed. We applied a machine learning model for explosion detection using smartphone microphones. The data used were from the Smartphone High-explosion Audio Recordings Dataset (SHAReD), a collection of 326 waveforms from 70 high-explosive (HE) events recorded on smartphones, and the ESC-50 dataset, a benchmarking dataset commonly used for environmental sound classification. Two machine learning models were trained and combined into an ensemble model for explosion detection. The resulting ensemble model classified audio signals as either “explosion,” “ambient,” or “other” with true positive rates (recall) greater than 96% for all three categories.
- 25 July 2024
- Published by MDPI AG in Preprints.org
Open AccessExplosion Detection using Smartphones: Ensemble Learning with the Smartphone High-explosive Audio Recordings Dataset and the ESC-50 Dataset
Explosion monitoring is performed by infrasound and seismoacoustic sensor networks that are distributed globally, regionally, and locally. However, these networks are unevenly and sparsely distributed, especially in the local scale as maintaining and deploying networks is costly. With increasing interest in smaller yield explosions, the need for more dense networks has increased. To address this issue, we propose using smartphone sensors for explosion detection as they are cost-effective and easy to deploy. Although there are studies using smartphone sensors for explosion detection, the field is still in its infancy and new technologies need to be developed. We applied a machine learning model for explosion detection using smartphone microphones. The data used were from the Smartphone High-explosion Audio Recordings Dataset (SHAReD), a collection of 326 waveforms from 70 high-explosive (HE) events recorded on smartphones, and the ESC-50 dataset, a benchmarking dataset commonly used for environmental sound classification. Two machine learning models were trained and combined into an ensemble model for explosion detection. The resulting ensemble model classified audio signals as either “explosion,” “ambient,” or “other” with true positive rates (recall) greater than 96% for all three categories.
- Open Access
- 19 October 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (20)
Acoustic Waves From a Distant Explosion Recorded on a Continuously Ascending Balloon in the Middle Stratosphere
A helium‐filled mylar balloon carrying a smartphone and infrasound sensors ascended to a stratospheric height of 35 km over the surface detonation of a chemical explosive, with a total acoustic propagation distance of 127 km. The smartphone was configured to collect multi‐modal data at high rates from internal sensors. Analysis of the data shows successful collection of the explosion signal by both the smartphone's microphone and its accelerometers, the first from an ascending balloon. Comparison of the acoustic signal with that collected by other infrasound sensors, both airborne and ground‐based, provides insight into the possibilities and limitations of collecting acoustic data from the stratosphere.
- 19 October 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (20)
Open AccessAcoustic Waves From a Distant Explosion Recorded on a Continuously Ascending Balloon in the Middle Stratosphere
A helium‐filled mylar balloon carrying a smartphone and infrasound sensors ascended to a stratospheric height of 35 km over the surface detonation of a chemical explosive, with a total acoustic propagation distance of 127 km. The smartphone was configured to collect multi‐modal data at high rates from internal sensors. Analysis of the data shows successful collection of the explosion signal by both the smartphone's microphone and its accelerometers, the first from an ascending balloon. Comparison of the acoustic signal with that collected by other infrasound sensors, both airborne and ground‐based, provides insight into the possibilities and limitations of collecting acoustic data from the stratosphere.
- Open Access
Laser-Induced Fluorescence for Monitoring Environmental Contamination and Stress in the Moss Thuidium plicatile
The ability to detect, measure, and locate the source of contaminants, especially heavy metals and radionuclides, is of ongoing interest. A common tool for contaminant identification and bioremediation is vegetation that can accumulate and indicate recent and historic pollution. However, large-scale sampling can be costly and labor-intensive. Hence, non-invasive in-situ techniques such as laser-induced fluorescence (LIF) are becoming useful and effective ways to observe the health of plants through the excitation of organic molecules, e.g., chlorophyll. The technique presented utilizes images collected of LIF in moss to identify different metals and environmental stressors. Analysis through image processing of LIF response was key to identifying Cu, Zn, Pb, and a mixture of the metals at nmol/cm2 levels. Specifically, the RGB values from each image were used to create density histograms of each color channel’s relative pixel abundance at each decimal code value. These histograms were then used to compare color shifts linked to the successful identification of contaminated moss samples. Photoperiod and extraneous environmental stressors had minimal impact on the histogram color shift compared to metals and presented with a response that differentiated them from metal contamination.
Open AccessLaser-Induced Fluorescence for Monitoring Environmental Contamination and Stress in the Moss Thuidium plicatile
The ability to detect, measure, and locate the source of contaminants, especially heavy metals and radionuclides, is of ongoing interest. A common tool for contaminant identification and bioremediation is vegetation that can accumulate and indicate recent and historic pollution. However, large-scale sampling can be costly and labor-intensive. Hence, non-invasive in-situ techniques such as laser-induced fluorescence (LIF) are becoming useful and effective ways to observe the health of plants through the excitation of organic molecules, e.g., chlorophyll. The technique presented utilizes images collected of LIF in moss to identify different metals and environmental stressors. Analysis through image processing of LIF response was key to identifying Cu, Zn, Pb, and a mixture of the metals at nmol/cm2 levels. Specifically, the RGB values from each image were used to create density histograms of each color channel’s relative pixel abundance at each decimal code value. These histograms were then used to compare color shifts linked to the successful identification of contaminated moss samples. Photoperiod and extraneous environmental stressors had minimal impact on the histogram color shift compared to metals and presented with a response that differentiated them from metal contamination.
- Open Access
An Assessment of Persistent Acoustic Monitoring of a Nuclear Reactor during Full Power Generation
Persistent low-frequency (<180 Hz) acoustic detection took place within the boundaries of Oak Ridge National Laboratory to monitor full power operations of the High Flux Isotope Reactor. Three acoustic sensors were installed at distances of 69, 101, and 914 m from the northeast corner of the cooling towers to monitor and assess four reactor power generation cycles. Features were extracted from power spectral density calculations where data were collected during reactor on and off operations. Diverse spectral features were present during full reactor power, including a 21.4 Hz fundamental frequency and ascending harmonics. Using bandpass filters, these related frequencies were isolated and summed, and the root mean square energy was calculated. The method of isolating and summing characteristic features provided a significant improvement in identifying acoustic behavior related to reactor power when the raw signals were obscured by noise.
Open AccessAn Assessment of Persistent Acoustic Monitoring of a Nuclear Reactor during Full Power Generation
Persistent low-frequency (<180 Hz) acoustic detection took place within the boundaries of Oak Ridge National Laboratory to monitor full power operations of the High Flux Isotope Reactor. Three acoustic sensors were installed at distances of 69, 101, and 914 m from the northeast corner of the cooling towers to monitor and assess four reactor power generation cycles. Features were extracted from power spectral density calculations where data were collected during reactor on and off operations. Diverse spectral features were present during full reactor power, including a 21.4 Hz fundamental frequency and ascending harmonics. Using bandpass filters, these related frequencies were isolated and summed, and the root mean square energy was calculated. The method of isolating and summing characteristic features provided a significant improvement in identifying acoustic behavior related to reactor power when the raw signals were obscured by noise.
- Open Access
Quantized Information in Spectral Cyberspace
The constant-Q Gabor atom is developed for spectral power, information, and uncertainty quantification from time–frequency representations. Stable multiresolution spectral entropy algorithms are constructed with continuous wavelet and Stockwell transforms. The recommended processing and scaling method will depend on the signature of interest, the desired information, and the acceptable levels of uncertainty of signal and noise features. Selected Lamb wave signatures and information spectra from the 2022 Tonga eruption are presented as representative case studies. Resilient transformations from physical to information metrics are provided for sensor-agnostic signal processing, pattern recognition, and machine learning applications.
Open AccessQuantized Information in Spectral Cyberspace
The constant-Q Gabor atom is developed for spectral power, information, and uncertainty quantification from time–frequency representations. Stable multiresolution spectral entropy algorithms are constructed with continuous wavelet and Stockwell transforms. The recommended processing and scaling method will depend on the signature of interest, the desired information, and the acceptable levels of uncertainty of signal and noise features. Selected Lamb wave signatures and information spectra from the 2022 Tonga eruption are presented as representative case studies. Resilient transformations from physical to information metrics are provided for sensor-agnostic signal processing, pattern recognition, and machine learning applications.
- Open Access
- 30 October 2022
- Published by Wiley in ESS Open Archive
Near-field source effects of the Tonga Lamb wave
A weather station in Nukuʻalofa (NUKU), Tonga, ~68km away from the epicenter of the 2022 Tonga eruption, recorded exceptional pressure, temperature, and wind data representative of the eruption source hydrodynamics. These high-quality data are available for further source and propagation studies. In contrast to other barometers and infrasound sensors at greater ranges, the NUKU barometer recorded a decrease in pressure during the climactic stage of the eruption. A simple fluid dynamic explanation of the depressurization is provided, with a commentary on near- vs far-field pressure observations of very large eruptions.
- 30 October 2022
- Published by Wiley in ESS Open Archive
Open AccessNear-field source effects of the Tonga Lamb wave
A weather station in Nukuʻalofa (NUKU), Tonga, ~68km away from the epicenter of the 2022 Tonga eruption, recorded exceptional pressure, temperature, and wind data representative of the eruption source hydrodynamics. These high-quality data are available for further source and propagation studies. In contrast to other barometers and infrasound sensors at greater ranges, the NUKU barometer recorded a decrease in pressure during the climactic stage of the eruption. A simple fluid dynamic explanation of the depressurization is provided, with a commentary on near- vs far-field pressure observations of very large eruptions.