Multitaper Spectral Estimation Research Articles

PO-04-137 SCAR TRANSMURALITY IN VENTRICULAR ENDOCARDIAL UNIPOLAR ELECTROGRAMS IS RELATED TO POWER OF PATIENT-SPECIFIC FREQUENCY BANDS

Characteristic endocardial electrogram (EGM) features such as fractionation and late potentials have been associated with ventricular tachycardia (VT) substrate. Our prior analyses show that EGM frequency content allows for more accurate identification of mid-myocardial fibrosis within a given patient, compared to traditional bipolar or unipolar voltage thresholds. A relationship between frequency information and endocardial scar transmurality has not been described. To explore the relationship of frequency spectra to ventricular scar transmurality in patients undergoing VT catheter ablation. Thirteen patients with non-ischemic cardiomyopathy (NICM) undergoing VT ablation with pre-procedural tissue characterization using cardiac computed tomography (CCT) imaging with late-iodine enhancement were assessed. From the CCT data, scar transmurality was calculated for every EGM recording location. With multitaper spectral estimation, average normalized frequency content was calculated for regions with different levels of scar transmurality and compared to average frequency content from tissue locations with no scar. K-nearest neighbors classifiers were trained within patients to predict scar transmurality, assuming presence of scar is known. Frequency content of regions without scar is similar among patients. A comparison of the average normalized frequency spectra of the non-fibrotic endocardial tissue for each patient is presented. Differences in scar transmurality led to changes in strength of certain frequency bands, which varied between patients. Observed responses included a peak at 10-15 Hz whose magnitude increased with scar transmurality (panel B) and a peak at 35-40 Hz in regions with higher scar transmurality. Normalized frequency content predicted transmural scar above chance with at least p < .05 for all 13 patients. Frequency content of endocardial unipolar EGMs is related to scar transmurality in those regions. This relationship changes patient to patient, potentially due to differences in morphology of fibrotic tissue. Deviation from the consistent spectra seen in non-fibrotic tissue may be a strategy to identify fibrosis and these insights can inform development of physiological models of VT. Because of low EGM sample sizes in higher scar transmurality regions, more data should be collected to further support this claim.

Spectral Connectivity: a python package for computing multitaper spectral estimates and frequency-domain brain connectivity measures on the CPU and GPU

Denovellis et al., (2022). Spectral Connectivity: a python package for computing multitaper spectral estimates and frequency-domain brain connectivity measures on the CPU and GPU. Journal of Open Source Software, 7(80), 4840, https://doi.org/10.21105/joss.04840

An adaptive multi-taper spectral estimation for stationary processes

The multi-taper spectral estimation method is effective in reducing both bias and variance of the spectrum of stationary stochastic processes. The number of tapers (NoT) in the method controls the bias and variance trade-off of the spectrum. In general, the NoT is empirically determined and constant at all frequencies, limiting the adjustment of the local bias and variance. This paper develops an adaptive multi-taper approach with the varying NoT at each frequency point for estimating the power spectral density (PSD) and coherence function of multivariate stationary processes. An iterative procedure with a stopping criterion is proposed to optimize the NoT at each frequency without manual tunning. The sine taper with a simple analytical expression and satisfactory leakage protection is employed. The proposed adaptive multi-taper approach is applied to three numerical examples, the structural response of a building model, a wind speed process, and an autoregressive process, which exhibit different spectral characteristics. The results of all examples show that the proposed approach outperforms Welch’s and traditional multi-taper methods in estimating the PSD and coherence with a smaller bias and variance.

Putting the Significance of Spectral Peaks on the Level: Implications for the 1470-Yr Peak in Greenland δ18O

Abstract Spectral analysis of the Greenland Ice Sheet Project 2 (GISP2) δ18O record has been interpreted to show a 1/(1470 yr) spectral peak that is highly statistically significant (p &lt; 0.01). The presence of such a peak, if accurate, provides an important clue about the mechanisms controlling glacial climate. As is standard, however, statistical significance was judged relative to a null model, H0, consisting of an autoregressive order one process, AR(1). In this study, H0 is generalized using an autoregressive moving-average process, ARMA(p, q). A rule of thumb is proposed for evaluating the adequacy of H0 involving comparing the expected and observed variances of the logarithm of a spectral estimate, which are generally consistent insomuch as removal of the ARMA structure from a time series results in an approximately level spectral estimate. An AR(1), or ARMA(1, 0), process is shown to be an inadequate representation of the GISP2 δ18O structure, whereas higher-order ARMA processes result in approximately level spectral estimates. After suitably leveling GISP2 δ18O and accounting for multiple hypothesis testing, multitaper spectral estimation indicates that the 1/(1470 yr) peak is insignificant. The seeming prominence of the 1/(1470 yr) peak is explained as the result of evaluating a spectrum involving higher-order ARMA structure and the peak having been selected on the basis of its seeming anomalous. The proposed technique for evaluating the significance of spectral peaks is also applicable to other geophysical records. Significance Statement A suitable null hypothesis is necessary for obtaining accurate test results, but a means for evaluating the adequacy of a null hypothesis for a spectral peak has been lacking. A generalized null model is presented in the form of an autoregressive, moving-average process whose adequacy can be gauged by comparing the observed and expected variance of log spectral density. Application of the method to the GISP2 δ18O record indicates that spectral structure found at 1/(1470 yr) is statistically insignificant.

Perioperative Electroencephalogram Spectral Dynamics Related to Postoperative Delirium in Older Patients.

Intraoperative electroencephalography (EEG) signatures related to the development of postoperative delirium (POD) in older patients are frequently studied. However, a broad analysis of the EEG dynamics including preoperative, postinduction, intraoperative and postoperative scenarios and its correlation to POD development is still lacking. We explored the relationship between perioperative EEG spectra-derived parameters and POD development, aiming to ascertain the diagnostic utility of these parameters to detect patients developing POD. Patients aged ≥65 years undergoing elective surgeries that were expected to last more than 60 minutes were included in this prospective, observational single center study (Biomarker Development for Postoperative Cognitive Impairment [BioCog] study). Frontal EEGs were recorded, starting before induction of anesthesia and lasting until recovery of consciousness. EEG data were analyzed based on raw EEG files and downloaded excel data files. We performed multitaper spectral analyses of relevant EEG epochs and further used multitaper spectral estimate to calculate a corresponding spectral parameter. POD assessments were performed twice daily up to the seventh postoperative day. Our primary aim was to analyze the relation between the perioperative spectral edge frequency (SEF) and the development of POD. Of the 237 included patients, 41 (17%) patients developed POD. The preoperative EEG in POD patients was associated with lower values in both SEF (POD 13.1 ± 4.6 Hz versus no postoperative delirium [NoPOD] 17.4 ± 6.9 Hz; P = .002) and corresponding γ-band power (POD -24.33 ± 2.8 dB versus NoPOD -17.9 ± 4.81 dB), as well as reduced postinduction absolute α-band power (POD -7.37 ± 4.52 dB versus NoPOD -5 ± 5.03 dB). The ratio of SEF from the preoperative to postinduction state (SEF ratio) was ~1 in POD patients, whereas NoPOD patients showed a SEF ratio >1, thus indicating a slowing of EEG with loss of unconscious. Preoperative SEF, preoperative γ-band power, and SEF ratio were independently associated with POD (P = .025; odds ratio [OR] = 0.892, 95% confidence interval [CI], 0.808-0.986; P = .029; OR = 0.568, 95% CI, 0.342-0.944; and P = .009; OR = 0.108, 95% CI, 0.021-0.568, respectively). Lower preoperative SEF, absence of slowing in EEG while transitioning from preoperative state to unconscious state, and lower EEG power in relevant frequency bands in both these states are related to POD development. These findings may suggest an underlying pathophysiology and might be used as EEG-based marker for early identification of patients at risk to develop POD.

Precise measurement of correlations between frequency coupling and visual task performance

Functional connectivity analyses focused on frequency-domain relationships, i.e. frequency coupling, powerfully reveal neurophysiology. Coherence is commonly used but neural activity does not follow its Gaussian assumption. The recently introduced mutual information in frequency (MIF) technique makes no model assumptions and measures non-Gaussian and nonlinear relationships. We develop a powerful MIF estimator optimized for correlating frequency coupling with task performance and other relevant task phenomena. In light of variance reduction afforded by multitaper spectral estimation, which is critical to precisely measuring such correlations, we propose a multitaper approach for MIF and compare its performance with coherence in simulations. Additionally, multitaper MIF and coherence are computed between macaque visual cortical recordings and their correlation with task performance is analyzed. Our multitaper MIF estimator produces low variance and performs better than all other estimators in simulated correlation analyses. Simulations further suggest that multitaper MIF captures more information than coherence. For the macaque data set, coherence and our new MIF estimator largely agree. Overall, we provide a new way to precisely estimate frequency coupling that sheds light on task performance and helps neuroscientists accurately capture correlations between coupling and task phenomena in general. Additionally, we make an MIF toolbox available for the first time.

Unified compressive sensing paradigm for the random demodulator and compressive multiplexer architectures

A major challenge in spectrum sensing for cognitive radio (CR) applications is the very high sampling rates involved, which imposes significant demands on the signal acquisition technology. This has given impetus to applying compressive sensing (CS) as a sub-Nyquist sampling paradigm for CR-type wireless signals which exhibit sparsity in certain domains. CS architectures like the random demodulator (RD) and compressive multiplexer (CM) can be used for CR spectral sensing, though both are inherently restricted in terms of the signal classes they can effectively process. To address these limitations, this paper presents two unified RD and CM-based CS architectures that seamlessly integrate precolouring and the multitaper spectral estimator into their respective structures to facilitate efficient sensing of both digitally modulated and narrowband signals, along with popular CR-access technologies like orthogonal frequency division multiplexing. A significant feature of these unified CS architectures is they do not require a priori knowledge of either the input signal or modulation scheme, while a tristate spectral classifier is introduced to afford notably enhanced spectrum access opportunities for unlicensed secondary users. A critical performance evaluation corroborates that both unified architectures demonstrate consistently superior CS results and robustness across a broad range of CR-type signals, modulations and access technologies.

GLRT-based spectrum sensing by exploiting Multitaper Spectral Estimation for cognitive radio network

Spectrum sensing is a key technique for future intelligent networks, such as AdHoc network, wireless sensor networks. However, the existing works suffer from high computational complexity, which is a huge challenge for those networks composed of many nodes with the limited computing capabilities. In this paper, we propose a simple and novel method based on the generalized likelihood ratio test(GLRT) criterion and offer a specific algorithm based on maximum value of power spectrum calculated by multitaper spectral estimation. We first derive the statistical characteristics and probability density function of test statistic, and the closed-form expression of detection probability and false-alarm probability are obtained. Then, we carry out some simulations to verify the proposed algorithm by taking BPSK signal as an example. In the simulations, we compare the sensing performance between the proposed algorithm and the existing methods. Both theoretical analysis and simulation results demonstrate the advantages of the proposed algorithm over the existing methods.

Dexmedetomidine-induced deep sedation mimics non-rapid eye movement stage 3 sleep: large-scale validation using machine learning.

Study ObjectivesDexmedetomidine-induced electroencephalogram (EEG) patterns during deep sedation are comparable with natural sleep patterns. Using large-scale EEG recordings and machine learning techniques, we investigated whether dexmedetomidine-induced deep sedation indeed mimics natural sleep patterns.MethodsWe used EEG recordings from three sources in this study: 8,707 overnight sleep EEG and 30 dexmedetomidine clinical trial EEG. Dexmedetomidine-induced sedation levels were assessed using the Modified Observer’s Assessment of Alertness/Sedation (MOAA/S) score. We extracted 22 spectral features from each EEG recording using a multitaper spectral estimation method. Elastic-net regularization method was used for feature selection. We compared the performance of several machine learning algorithms (logistic regression, support vector machine, and random forest), trained on individual sleep stages, to predict different levels of the MOAA/S sedation state.ResultsThe random forest algorithm trained on non-rapid eye movement stage 3 (N3) predicted dexmedetomidine-induced deep sedation (MOAA/S = 0) with area under the receiver operator characteristics curve >0.8 outperforming other machine learning models. Power in the delta band (0–4 Hz) was selected as an important feature for prediction in addition to power in theta (4–8 Hz) and beta (16–30 Hz) bands.ConclusionsUsing a large-scale EEG data-driven approach and machine learning framework, we show that dexmedetomidine-induced deep sedation state mimics N3 sleep EEG patterns.Clinical TrialsName—Pharmacodynamic Interaction of REMI and DMED (PIRAD), URL—https://clinicaltrials.gov/ct2/show/NCT03143972, and registration—NCT03143972.

0449 Characterizing Spindle Activity as a Time-Frequency Phenomenon

Abstract Introduction Spindles are currently defined clinically based on observed patterns in the EEG waveform trace, with automated methods seeking to replicate visual scoring by experts. Recent work suggests that sleep spindles may be more readily observed as time-frequency peaks in the EEG spectrogram. This study compares spectral peaks in the multitaper spectrogram to expert and automatic detection scoring, characterizes the variability of spindles across a night, and investigates topographical and temporal clustering of spindles within individual EEG records. Methods We compared spectral peaks, expert scoring, and automatic detection in two datasets (DREAMS, and a high-density control study). Peaks were identified using multitaper spectral estimation and the peak prominence of the normalized power spectrum for each channel. Spatiotemporal variability analysis was performed using cluster and pattern recognition algorithms including penalized sorting of channel activation order, 2D-cross correlation, PCA and UMAP cluster analysis, and the seqNMF method. Results Spectral peaks were shown to be highly robust to and easily differentiated from broadband noise, occuring at rates (10-16 per min) far exceeding spindle rates reported in literature (~2.5 per min). Expert scoring and automated scoring failed to capture clear spectral peaks in the time-frequency domain, indicating an underreporting of the phenomenology. No apparent clustering or patterns of sleep spindle-like activity was observed using the proposed methods, suggesting high variability of spatiotemporal evolution of spindles. Conclusion These results suggest that the difficulty of time-domain visual scoring of spindles causes an artificially low estimate of the underlying phenomenology, which is mirrored in the assumptions implicit in the thresholds of automated scorers. This work shows that spindles are highly variable in their spatiotemporal evolution, suggesting that there is no optimal single electrode for analysis and casting doubt on the presence of a single cortical generation mechanism. We must therefore revisit the concept of the spindle using the time-frequency domain to more robustly characterize underlying phenomenology. Support National Institute Of Neurological Disorders And Stroke Grant R01 NS-096177

Improving the Lomb–Scargle Periodogram with the Thomson Multitaper

Abstract A common approach for characterizing the properties of time-series data that are evenly sampled in time is to estimate the power spectrum of the data using the periodogram. The periodogram as an estimator of the spectrum is (1) statistically inconsistent (i.e., its variance does not go to zero as infinite data are collected), (2) biased for finite samples, and (3) suffers from spectral leakage. In astronomy, time-series data are often unevenly sampled in time, and it is popular to use the Lomb–Scargle (LS) periodogram to estimate the spectrum. Unfortunately, from a statistical standpoint, the LS periodogram suffers from the same issues as the classical periodogram and has even worse spectral leakage. Here, we present an improvement on the LS periodogram by combining it with the Thomson multitaper approach. The multitaper spectral estimator is well established in the statistics and engineering literature for evenly sampled time series. It is attractive because it directly trades off bias and variance for frequency resolution, and is fast to compute: compared to an untapered spectral estimator, the multitaper adds no more than a couple of seconds for a time series with a million data points on a current desktop computer. Here, we describe an estimator that combines the multitaper with the LS periodogram. We show examples in which this new approach has improved properties compared to traditional approaches in the case of unevenly sampled time series. Finally, we demonstrate an application of the method to astronomy with an application to Kepler data.

BLSBA‐ED‐SSO: Bivariate Lévy‐stable bat algorithm‐based energy detection for spectrum sensing optimization in cognitive radio networks (CRNs)

SummaryThe vacant licensed spectrum can be employed by the secondary users in cognitive radio networks. Nevertheless, this process of identification is frequently agreed with shadowing, multipath fading and receiver uncertainty problems. The primary contributions of the spectrum sensing techniques adopted in this work mainly concentrate on energy detection for spectrum sensing as well as to recognize the optimal spectral estimation according to the multitaper spectral estimation. The algorithm proposed in this work is based on bivariate Lévy‐stable bat algorithm (BLSBA), energy detector (ED), and with the help of K out of M fusion rule; it is analysed for the single user as well as cooperative multiple users. In the BLSBA algorithm, a modified search equation with more helpful information from the search experiences is brought‐in to create an optimal energy solution and bivariate Lévy‐stable random walk is associated with BLSBA to remove the trapping process into local optima. The energy detection is defined numerically from this optimal detection. At last, a multi‐taper spectral estimator (MSE) is proposed to cognitive radio detection for a huge network. The simulation results in both cases are computed and checked with the help of a BLSBA optimizer. For an individual secondary user scenario, the advancement of MSE to the ED is broadly described. Experimental result indicates that the objective false alarm likelihood is minimized and the demanded signal‐to‐noise ratio is accomplished to the extent.

Spectral Estimation Using Multitaper Whittle Methods with a Lasso Penalty.

Spectral estimation provides key insights into the frequency domain characteristics of a time series. Naive non-parametric estimates of the spectral density, such as the periodogram, are inconsistent, and the more advanced lag window or multitaper estimators are often still too noisy. We propose an L 1 penalized quasi-likelihood Whittle framework based on multitaper spectral estimates which performs semiparametric spectral estimation for regularly sampled univariate stationary time series. Our new approach circumvents the problematic Gaussianity assumption required by least square approaches and achieves sparsity for a wide variety of basis functions. We present an alternating direction method of multipliers (ADMM) algorithm to efficiently solve the optimization problem, and develop universal threshold and generalized information criterion (GIC) strategies for efficient tuning parameter selection that outperform cross-validation methods. Theoretically, a fast convergence rate for the proposed spectral estimator is established. We demonstrate the utility of our methodology on simulated series and to the spectral analysis of electroencephalogram (EEG) data.

Efficient Recursive Least Square Technique for Spectrum Sensing in Cognitive Radio Networks

Cognitive radio-based systems rely on spectrum sensing techniques to detect whitespaces to exploit. Cognitive radio (CR) is an attractive approach to face the shortage in the electromagnetic spectrum resources and improve the overall spectrum utilization. However, Energy detectors perform far from optimally by affecting the performance of the underlying system. In this article, two spectrum-sensing techniques are considered for CR networks; one based on energy detection and the other based on multi-taper spectral estimation (MSE). This article proposes a new method to optimize the overall performance in cooperative spectrum sensing in cognitive radio (CR) networks. An efficient recursive least square (ERLS)-based approach is proposed in order to optimize the overall performance to monitor the primary user active or inactive stage with use of secondary user while receiving data. An energy detector (ED) and multi-taper (MTM) spectrum sensing techniques are examined as local spectrum sensing techniques. Finally, a genetic algorithm is compared with the proposed system to show the system effectiveness.

Multitaper Infinite Hidden Markov Model for EEG.

Electroencephalographam (EEG) monitoring of neural activity is widely used for identifying underlying brain states. For inference of brain states, researchers have often used Hidden Markov Models (HMM) with a fixed number of hidden states and an observation model linking the temporal dynamics embedded in EEG to the hidden states. The use of fixed states may be limiting, in that 1) pre-defined states might not capture the heterogeneous neural dynamics across individuals and 2) the oscillatory dynamics of the neural activity are not directly modeled. To this end, we use a Hierarchical Dirichlet Process Hidden Markov Model (HDP-HMM), which discovers the set of hidden states that best describes the EEG data, without a-priori specification of state number. In addition, we introduce an observation model based on classical asymptotic results of frequency domain properties of stationary time series, along with the description of the conditional distributions for Gibbs sampler inference. We then combine this with multitaper spectral estimation to reduce the variance of the spectral estimates. By applying our method to simulated data inspired by sleep EEG, we arrive at two main results: 1) the algorithm faithfully recovers the spectral characteristics of the true states, as well as the right number of states and 2) the incorporation of the multitaper framework produces a more stable estimate than traditional periodogram spectral estimates.

Separable Estimation of Ambient Noise Spectrum in Synchrophasor Measurements in the Presence of Forced Oscillations

The ambient noise in synchrophasor measurements carries core dynamic signatures of a power system and its spectrum can strongly reflect the dynamic properties and operating conditions of the system and its loads. Typically, the ambient noise spectrum is estimated from a window of measurements captured during steady state operating conditions of the system. The presence of any Forced Oscillation (FO) in the measurement window severely biases the estimated ambient noise spectrum. In this article, an algorithm is proposed for separable estimation of the ambient noise spectrum in the presence of periodic FOs in synchrophasor measurements. The algorithm utilizes the Thomson's multitaper spectral estimation and harmonic analysis techniques and is capable of effectively reducing the bias in estimating the ambient noise spectrum due to the presence of FOs in the measurements. The performance of the estimator is analyzed theoretically and the theoretical results are verified experimentally using simulation studies. Application of the algorithm to field measured data illustrates that the algorithm is capable of effectively reducing the bias introduced by FOs in estimating the ambient noise spectrum from the synchrophasor measurements containing both ambient noise and FOs.

Multitaper Analysis of Semi-Stationary Spectra from Multivariate Neuronal Spiking Observations.

Extracting the spectral representations of neural processes that underlie spiking activity is key to understanding how brain rhythms mediate cognitive functions. While spectral estimation of continuous time-series is well studied, inferring the spectral representation of latent non-stationary processes based on spiking observations is challenging due to the underlying nonlinearities that limit the spectrotemporal resolution of existing methods. In this paper, we address this issue by developing a multitaper spectral estimation methodology that can be directly applied to multivariate spiking observations in order to extract the semi-stationary spectral density of the latent non-stationary processes that govern spiking activity. We establish theoretical bounds on the bias-variance trade-off of our proposed estimator. Finally, application of our proposed technique to simulated and real data reveals significant performance gains over existing methods.

A multitaper spectral estimator for time-series with missing data

SUMMARY A multitaper estimator is proposed that accommodates time-series containing gaps without using any form of interpolation. In contrast with prior missing-data multitaper estimators that force standard Slepian sequences to be zero at gaps, the proposed missing-data Slepian sequences are defined only where data are present. The missing-data Slepian sequences are frequency independent, as are the eigenvalues that define the energy concentration within the resolution bandwidth, when the process bandwidth is $[ { - 1/2,\,\,\,1/2} )$ for unit sampling and the sampling scheme comprises integer multiples of unity. As a consequence, one need only compute the ensuing missing-data Slepian sequences for a given sampling scheme once, and then the spectrum at an arbitrary set of frequencies can be computed using them. It is also shown that the resulting missing-data multitaper estimator can incorporate all of the optimality features (i.e. adaptive-weighting, F-test and reshaping) of the standard multitaper estimator, and can be applied to bivariate or multivariate situations in similar ways. Performance of the missing-data multitaper estimator is illustrated using length of day, seafloor pressure and Nile River low stand time-series.

Detection of Periodic Forced Oscillations in Power Systems Using Multitaper Approach

An algorithm based on Thomson's multitaper spectral estimation and harmonic analysis techniques is proposed for the detection and frequency estimation of periodic forced oscillations in a power system. Unlike traditional periodogram-based techniques, the method does not rely on true ambient noise spectrum of the system, and can operate continuously in an online environment without requiring any system knowledge except synchrophasor measurements. It compares a test statistic derived from only measurements against a threshold, and the threshold is established from general expressions of the test statistic's distribution and the probability of false alarm. Performance of the detector is explicitly expressed using probability of detection expression, and evaluated using simulation studies. Results upon application of this algorithm to field measured synchrophasor data suggest usability of the technique for forced oscillation detection in practical monitoring of the power system.

High‐Q Spectral Peaks and Nonstationarity in the Deep Ocean Infragravity Wave Band: Tidal Harmonics and Solar Normal Modes

AbstractInfragravity waves have received the least study of any class of waves in the deep ocean. This paper analyzes a 389‐day‐long deep ocean pressure record from the Hawaii Ocean Mixing Experiment for the presence of narrowband (≲2 μHz) components and nonstationarity over 400–4,000 μHz using a combination of fitting a mixture noncentral/central χ2 model to spectral estimates, high‐resolution multitaper spectral estimation, and computation of the offset coherence between distinct frequencies for a given data segment. In the frequency band 400–1,000 μHz there is a noncentral fraction of 0.67 ± 0.07 that decreases with increasing frequency. Evidence is presented for the presence of tidal harmonics in the data over the 400‐ to 1,400‐μHz bands. Above ~2,000 μHz the noncentral fraction rises with frequency, comprising about one third of the spectral estimates over 3,000–4,000 μHz. The power spectrum exhibits frequent narrowband peaks at 6–11 standard deviations above the noise level. The widths of the peaks correspond to a Q of at least 1,000, vastly exceeding that of any oceanic or atmospheric process. The offset coherence shows that the spectral peaks have substantial (p = 0.99–0.9999) interfrequency correlation, both locally and between distinct peaks within a given analysis band. Many of the peak frequencies correspond to the known values for solar pressure modes that have previously been observed in solar wind and terrestrial data, while others are the result of nonstationarity that distributes power across frequency. Overall, this paper documents the existence of two previously unrecognized sources of infragravity wave variability in the deep ocean.