Total Frequency Band Research Articles

MDD brain network analysis based on EEG functional connectivity and graph theory

BackgroundExisting studies have shown that the brain network of major depression disorder (MDD) has abnormal topologies. However, constructing reliable MDD brain networks is still an open problem. New methodThis paper proposed a reliable MDD brain network construction method. First, seven connectivity methods are used to calculate the correlation between channels and obtain the functional connectivity matrix. Then, the matrix is binarized using four binarization methods to obtain the EEG brain network. Besides, we proposed an improved binarization method based on the criterion of maximizing differences between groups: the adaptive threshold (AT) method. The AT can automatically set the optimal binarization threshold and overcome the artificial influence of traditional methods. After that, several network metrics are extracted from the brain network to analyze inter-group differences. Finally, we used statistical analysis and Fscore values to compare the performance of different methods and establish the most reliable method for brain network construction. ResultsIn theta, alpha, and total frequency bands, the clustering coefficient, global efficiency, local efficiency, and degree of the MDD brain network decrease, and the path length of the MDD brain network increases. Comparison with existing methodsThe results show that AT outperforms the existing binarization methods. Compared with other methods, the brain network construction method based on phase-locked value (PLV) and AT has better reliability. ConclusionsMDD has brain dysfunction, particularly in the frontal and temporal lobes.

Microwave heating characteristics of cement mortar containing carbonyl iron powder applied to airport pavement deicing

To improve the microwave deicing efficiency of airport concrete pavement, carbonyl iron powder (CIP) was used as an absorbing material to prepare the microwave-absorbing cement mortar. Firstly, to characterize the microwave absorption ability of the mortar, its electromagnetic (EM) parameters and microwave reflection loss (RL) with different CIP content were investigated by using the coaxial vector network analyzer (CVNA) and arching reflected method, respectively. Then, the microwave heating rate and deicing efficiency of the mortar coated with ice was investigated under different icing temperature and ice layer thickness conditions with a microwave heating frequency of 2.45 GHz. Finally, the mechanical strength of mortar after deicing was tested to analyze the effect of an icing-deicing cycle on its engineering performance. The results showed that at the total frequency band, the mean RL and effective bandwidth of 10CCM was −2.8 dB lower and 1.6 GHz wider than that of the control group. Meanwhile, at 2.45 GHz, the RL of 10CCM was 1.31 times lower than that of the control group. When the icing temperature and ice layer thickness were − 10 °C and 1 cm, the heating rate of 10CCM was 1.34 times higher than that of the control group. Compared to the icing temperature, the changes in ice layer thickness on the mortar surface had less effect on its heating rate. After deicing, the CIP mortar still had favorable mechanical strength. Hence, the CIP was a favorable absorbing material that could be applied to cement mortar to improve the microwave deicing efficiency of the concrete pavement during long-term service.

Clinical Features of Non-Lateralized Tinnitus.

We aimed to analyze and confirm the clinical features of patients with non-lateralized tinnitus and to identify clues that can be used in their management. Data from 469 patients who visited a university hospital complaining of tinnitus between March 2020 and December 2021 were reviewed. The patients' medical histories, Tinnitus Handicap Inventory, Beck Depression Inventory, and numerical rating scale scores on tinnitus awareness, annoyance, loudness, and effect on life, audiological profiles, and quantitative electroencephalography findings were documented. Forty-nine (10.4%) patients had non-lateralized tinnitus. They were older and had a shorter duration of symptoms (13.91 ± 34.16 months) than patients with bilateral tinnitus (duration: 39.15 ± 80.82 months) (P -lt; .05). The accompanying symptoms, Tinnitus Handicap Inventory scores, and numerical rating scale scores were not significantly different between the 2 groups (P -gt; .05). Patients with non-lateralized tinnitus had worse hearing at 12 kHz on the left side than those with unilateral tinnitus. Hearing asymmetry was least common in non-lateralized tinnitus (n=11/49, 10.4%), followed by bilateral tinnitus (n=54/198, 42.2%) and unilateral tinnitus (n=97/222, 47.3%) (P-lt; .001). Regarding quantitative electroencephalography, there were significant differences in the absolute power of the theta, alpha, beta, gamma, and total frequency bands based on tinnitus lateralization (P -lt; .001). Non-lateralized tinnitus can be perceived in elderly patients with symmetric and extended high-frequency hearing loss before habituation is achieved at an early stage of tinnitus. However, there was no difference in the questionnaire scores and accompanying symptoms; therefore, it may not be worth managing non-lateralized tinnitus separately from tinnitus in the ear.

Modeling D2D Underlaying Cellular Network for Hotspot Communications with Poisson Cluster and Hole Processes

This paper develops a new approach to the modeling and analysis of device-to-device (D2D) underlaying multi-tier cellular network for dense hotspot communications, which consists of macro base stations (MBSs), pico BSs (PBSs), femto BSs (FBSs). A typicl user equipment (UE) can work either in D2D mode or cellular mode. Considering the dense hotspot communications, this work employs Poisson point process (PPP) to model the locations of MBSs and PBSs, and uses Poisson cluster process (PCP) to model the ones of UEs and FBSs. The locations of PBSs are also modeled as the centers of hotspots, referred to as the centers of PCPs. UEs and FBSs cluster around the common parent process PBSs. To guard the cluster-edge UEs, the clustered-UE classification and modified fractional frequency reuse (FFR) are jointly used, by which both the UEs and FBSs are classified two sets, cluster-center UEs and cluster-edge UEs, cluster-center FBSs and cluster-edge FBSs, respectively. The total frequency band is divided into two orthogonal segments, one of which is shared by D2D devices, cluster-edge FBSs, and PBSs, and the other segment of which is shared by cluster-center FBSs and MBSs. For such clustered multi-tier network, by using the methods from PPP, PCP, and PHP, this paper presents a tractable approach for modeling and analyzing the performance of cellular and D2D networks and gives the statistical descriptions of the experienced interferences at a typical D2D or cellular receiver by using the approximated Poisson hole processes (PHP) theory. This yields the derivations of the coverage probabilities of both the D2D receivers and cellular destinations. In additon, during the analysis of cellular UEs, to derive the coverage probabilities, this paper specially constructs one UE association criterion as well as the derivations of both the association probabilities and the statistical descriptions of association distances for cluster-center and cluster-edge UEs. The simulations results exploit the effect of various network parameters on the network performance and give the insights in terms of the proposed schemes as well as the comparison between cluster-center and cluster-edge UEs.

EEG-Based Emotion Recognition Using Logistic Regression with Gaussian Kernel and Laplacian Prior and Investigation of Critical Frequency Bands

Emotion plays a nuclear part in human attention, decision-making, and communication. Electroencephalogram (EEG)-based emotion recognition has developed a lot due to the application of Brain-Computer Interface (BCI) and its effectiveness compared to body expressions and other physiological signals. Despite significant progress in affective computing, emotion recognition is still an unexplored problem. This paper introduced Logistic Regression (LR) with Gaussian kernel and Laplacian prior for EEG-based emotion recognition. The Gaussian kernel enhances the EEG data separability in the transformed space. The Laplacian prior promotes the sparsity of learned LR regressors to avoid over-specification. The LR regressors are optimized using the logistic regression via variable splitting and augmented Lagrangian (LORSAL) algorithm. For simplicity, the introduced method is noted as LORSAL. Experiments were conducted on the dataset for emotion analysis using EEG, physiological and video signals (DEAP). Various spectral features and features by combining electrodes (power spectral density (PSD), differential entropy (DE), differential asymmetry (DASM), rational asymmetry (RASM), and differential caudality (DCAU)) were extracted from different frequency bands (Delta, Theta, Alpha, Beta, Gamma, and Total) with EEG signals. The Naive Bayes (NB), support vector machine (SVM), linear LR with L1-regularization (LR_L1), linear LR with L2-regularization (LR_L2) were used for comparison in the binary emotion classification for valence and arousal. LORSAL obtained the best classification accuracies (77.17% and 77.03% for valence and arousal, respectively) on the DE features extracted from total frequency bands. This paper also investigates the critical frequency bands in emotion recognition. The experimental results showed the superiority of Gamma and Beta bands in classifying emotions. It was presented that DE was the most informative and DASM and DCAU had lower computational complexity with relatively ideal accuracies. An analysis of LORSAL and the recently deep learning (DL) methods is included in the discussion. Conclusions and future work are presented in the final section.

Depression recognition using machine learning methods with different feature generation strategies.

The diagnosis of depression almost exclusively depends on doctor-patient communication and scale analysis, which have the obvious disadvantages such as patient denial, poor sensitivity, subjective biases and inaccuracy. An objective, automated method that predicts clinical outcomes in depression is essential for increasing the accuracy of depression recognition and treatments. This paper aims at better recognizing depression using the transformation of EEG features and machine learning methods. An experiment based on emotional face stimuli task was conducted, and twenty-eight subjects' EEG data were recorded from 128-channel HydroCel Geodesic Sensor Net (HCGSN) by Net Station software. The Mini International Neuropsychiatric Interview (MINI) was used by psychiatrists as the criterion for diagnosis of depression patients. The power spectral density and activity were respectively extracted as original features using Auto-regress model and Hjorth algorithm with different time windows. Two separate approaches processed the features: ensemble learning and deep learning. For the ensemble learning, a deep forest transformed the original features to new features that potentially improve feature engineering and a support vector machine (SVM) that was applied as classifier. For deep learning method, we added spatial information of EEG caps to both features by image conversion and adopted convolutional neural network (CNN) to recognize them. The performance of both methods was evaluated for separated and total frequency bands. As a result, the best accuracy obtained was 89.02% when we used the ensemble model and power spectral density. The best accuracy of deep learning method was 84.75% using the activity. These experimental results prove the efficiency of the proposed methods and show that EEG could be used as a reliable indicator for depression recognition, which makes it possible for EEG-based portable system design and application in auxiliary depression recognition in the future.

The moderating effect of heart rate variability on the relationship between alpha asymmetry and depressive symptoms

Electroencephalographic (EEG) research has suggested relatively reduced brain activity in the left frontal and right posterior region trait-markers of depression. However, inconsistent results have been reported. Based on previous studies reporting the heart rate variability (HRV) as an index of emotional regulation, this study makes a novel investigation of the role of heart rate variability (HRV) as a moderator in the relationship between frontal and parietal alpha asymmetry and depression. Resting EEG (eyes open) was recorded in 38 patients with MDD and 34 healthy subjects. Frontal and parietal alpha asymmetries were calculated at total (8–12 Hz), high (10–12 Hz), and low (8–10 Hz) alpha frequency bands. Three vagally mediated HRV (vmHRV) components (LF, HF, and the LF/HF ratio) were calculated in the frequency domain. Relatively greater right parietal alpha activity significantly predicted the severity of depression only when HF was low (or the LF/HF ratio was high) at low alpha frequency band. The interaction effect of parietal alpha asymmetry and vmHRV remained significant after including anxiety score as a covariate. No moderation effect of vmHRV was found for frontal sites and other frequency bands, as well as healthy subjects. These findings suggest that vmHRV moderates the association between parietal alpha asymmetry at low frequency band and depression for MDD patients. We suggest that the interaction between parietal alpha asymmetry and vmHRV may be a biomarker of MDD.

Heart rate variability is associated with outcome in spontaneous intracerebral hemorrhage

PurposeAutonomic imbalance as measured by heart rate variability (HRV) has been associated with poor outcome after stroke. Observations on HRV changes in intracerebral hemorrhage (ICH) are scarce. Here, we aimed to investigate HRV in ICH as compared to a control group and to explore associations with stroke severity, hemorrhage volume and outcome after ICH. MethodsWe examined the autonomic modulation using frequency domain analysis of HRV during the acute phase of the ICH and in a healthy age- and hypertension-matched control group. Hematoma volume, intraventricular extension, initial stroke severity and baseline demographic, clinical parameters as well as mortality and functional outcome were included in the analysis. Results47 patients with ICH and 47 age- and hypertension matched controls were analyzed. ICH patients showed significantly lower total high frequency band (HF) and low frequency band (LF) powers (p = 0.01, p < 0.001), higher normalized HF power (p = 0.03), and lower LF/HF ratio (p < 0.001) as compared to the controls. Autonomic parameters showed associations with stroke severity (p = 0.004) and intraventricular involvement (p = 0.01) and predicted poor outcome independently (p = 0.02). ConclusionsAutonomic changes seems to be present in acute ICH and are associated with poor outcome independently. This may have future monitoring and therapeutic implications.

Parameterization of variations for wave energy over bottom slopes

Wave energy transfer through a shallow water region, produced by nonlinearity, was investigated using random wave simulations based on JONSWAP spectra in a physical experiment. To confirm the role of topography, slopes of β = 1/15, 1/30, and 1/45 were installed in a flume. The shoaling effect was first examined by comparing the experimental data with the theoretically predicted values. A wavelet-based bispectrum was then used to investigate the role of the bottom slope in energy transfer. Next, the evolution of energy transfer in the total, primary, and higher harmonic frequency bands was investigated using the wavelet-based energy spectrum. Based on the data from each slope, empirical formulas expressing the dimensionless energy parameters as a function of the local water depth were derived. The bottom slope was confirmed to have an influence on the variation in the total energy and primary harmonic energy, but its effect was shown to be negligible for the higher harmonic energy. Novel formulas were derived to represent the combined slope effect.

Investigating the effects of nitrous oxide sedation on frontal-parietal interactions

Although functional connectivity has received considerable attention in the study of consciousness, few studies have investigated functional connectivity limited to the sedated state where consciousness is maintained but impaired. The aim of the present study was to investigate changes in functional connectivity of the parietal-frontal network resulting from nitrous oxide-induced sedation, and to determine the neural correlates of cognitive impairment during consciousness transition states. Electroencephalography was acquired from healthy adult patients who underwent nitrous oxide inhalation to induce cognitive impairment, and was analyzed using Granger causality (GC). Periods of awake, sedation and recovery for GC between frontal and parietal areas in the delta, theta, alpha, beta, gamma and total frequency bands were obtained. The Friedman test with post-hoc analysis was conducted for GC values of each period for comparison. As a sedated state was induced by nitrous oxide inhalation, power in the low frequency band showed increased activity in frontal regions that was reversed with discontinuation of nitrous oxide. Feedback and feedforward connections analyzed in spectral GC were changed differently in accordance with EEG frequency bands in the sedated state by nitrous oxide administration. Calculated spectral GC of the theta, alpha, and beta frequency regions in the parietal-to-frontal direction was significantly decreased in the sedated state while spectral GC in the reverse direction did not show significant change. Frontal-parietal functional connectivity is significantly affected by nitrous oxide inhalation. Significantly decreased parietal-to-frontal interaction may induce a sedated state.

Inverse problem of the event reconstruction in the LORD experiment using the signal speсtrum information

Using the modified random search method, the solution of the inverse problem of reconstructing events in the LORDlunar experiment is considered using the signal spectral analysis, where the total frequency band of 200–400 MHz is divided into four equal frequency bands. Additional information about the signal will allows one to achieve a higher accuracy in the solution of the inverse problem.

제한된 주파수 대역에서 Link-16 웨이브폼의 멀티넷 성능 향상을 위한 그룹 기반의 주파수 도약 방식

Link-16 is a representative TDL operated by US air force and NATO and supports structure of Multi-net. Under Multi-net, military operation can be conducted effectively since terminal nodes in Link-16 hop over total frequency band simultaneously. As air traffic is rapidly increasing, new aeronautical system is introduced or existing system should be expanded to accommodate increasing air traffic and frequency band assigned for operating this system is scarce. It is scheduled to implement frequency remapping to solve frequency scarcity. With limited frequency band for operating Link-16, as frequency remapping is implemented, degradation of Multi-net performance can happen since multiple access interference in Link-16 is increasing so it is difficult to conduct multiple military operations. Thus, Group-based frequency hopping scheme is proposed to solve this problem. We verified the performance of the proposed scheme is improved.

EEG maturation and stability of cerebral oxygen extraction in very low birth weight infants.

Fractional cerebral tissue oxygen extraction (FTOE) can be continuously monitored by simultaneous near-infrared spectroscopy (NIRS) and pulse oximetry. The objective of this study is to test the hypothesis that in very low birth weight (VLBW) infants, the more mature EEG activity is, the less variable FTOE is. A prospective study was conducted on VLBW infants (< 1500 g and ⩽ 34 weeks gestation) without significant brain injury. Simultaneous continuous two-channel electroencephalography (EEG), NIRS and pulse oximetry were recorded. Absolute and relative powers of EEG in the delta, theta, alpha, beta and total frequency bands have been calculated. FTOE variability was calculated on two scales: short scales (3 to 20 s) and long scales (20 to 150 s). FTOE variability was examined against changes in relative spectral power of different EEG bands. We evaluated 67 studies performed on 46 VLBW infants. Average study duration was 21.3 ± 5.5 h. Relative power of delta band positively correlated with FTOE short- and long-scale variability (r=0.45, P<0.001; r=0.44, P<0.001, respectively). Relative power of alpha bands negatively correlated with FTOE short- and long-scale variability (r=-0.38, P=0.002; r=-0.42, P<0.001, respectively). These correlations continued to be significant when controlling for sex, small for gestational age, postmenstrual age, being on respiratory support, hemoglobin concentration, systemic oxygen saturation and transcutaneous carbon dioxide tension. Increased maturation of EEG activity is associated with decreased variability in cerebral oxygen extraction. The implications of increased variability in FTOE on brain injury in premature infants need further exploration.

Exposure to excessive sounds during orchestra rehearsals and temporary hearing changes in hearing among musicians.

It has been shown that musicians are at risk of noise-induced hearing loss. The aim of the study has been to evaluate the temporary changes of hearing in the case of orchestral musicians after group rehearsals. The study group comprised 18 orchestral musicians, aged 30-58 years old (mean: 40 years old) having 12-40 years (mean: 22 years) of professional experience. The temporary changes in hearing after group rehearsals were determined using transient-evoked otoacoustic emissions (TEOAEs). Noise exposures during group rehearsals were also evaluated. Musicians' hearing threshold levels were higher (worse) than expected for the equivalent non-noise-exposed population. Moreover, the high frequency notched audiograms were observed in some of them. After rehearsals, during which musicians were exposed to orchestral noise at A-weighted equivalent-continuous sound pressure level (normalized to 8-h working day) varied from 75.6-83.1 dB (mean: 79.4 dB). The significant post-exposure reductions of TEOAE amplitudes (approx. 0.7 dB) both for the total response and frequency bands of 2000 and 3000 Hz were noted. However, there were no significant differences between pre- and postexposure reproducibility of TEOAE. Obtained results have confirmed that orchestral musicians are at risk of hearing loss due to their professional activities, even at exposures to orchestral noise less than the limit values for occupational noise.

Wideband Spectrum Sensing for Cognitive Radios: A Multistage Wiener Filter Perspective

A wideband spectrum sensing scheme for cognitive radios is put forward in this letter. The proposed method operates over the total frequency bands simultaneously rather than a single band each time. It first estimates the number of occupied subbands, and then determines the accurate locations of occupied subbands as well as the vacant ones. Applying the ideas of multi stage Wiener filter to Gerschgorin disk estimator, this approach jointly makes the decision. In this way, the detection performance is enhanced because of capturing the signal information and suppressing the additive noise. Meanwhile, this approach avoids estimating the covariance matrix as well as the eigenvalue decomposition, and thus achieves a low computational complexity. Distinct from the conventional algorithms, neither noise power estimation nor prior knowledge of primary user signal is required, making the proposed algorithm robust to noise uncertainty and suitable for blind detection. Simulations under various conditions are presented to demonstrate the benefits of this approach and the results indicate that it surpasses the existing sensing algorithms.

Multiband Detection for Spectrum Sensing: A Multistage Wiener Filter Perspective

In the recent years, spectrum scarcity becomes an urgent issue due to the emergence of wireless services. The effective utilization of spectrum white space has gained significant research interests. Cognitive radio techniques have been paid much attention to the television white space. This paper raises a multiband spectrum sensing scheme to detect the spectrum white space which is not limited to television bands. When performing spectrum sensing, our approach operates over the total frequency bands simultaneously rather than a single band each time. By applying the idea of multistage Wiener filter to Gerschgorin disk estimator, our approach jointly makes the decision. In this way, the proposed method is able to capture the signal information and suppress the additive noise, which brings about an enhanced detection performance. Distinct from the classical methods, the proposed scheme requires neither noise power estimation nor prior knowledge of primary user signal, thereby being robust to noise uncertainty and suitable for blind detection. On the contrary, in the context of noise uncertainty, noise variance has no access to accurate estimation, inducing an imprecise decision threshold, which severely deteriorates the detection performance. Besides, our method avoids the estimation of covariance matrix as well as eigenvalue decomposition, and thus achieves a low computational complexity. This paper presents simulations under various conditions to verify the performance of the proposed scheme and the results show that it is superior to the existing sensing algorithms.

Forming OFDM symbols for 4G service signals with a total working frequency band of more than 3–4 GHz

The basis of the solution of the problem of forming OFDM symbols with a set of 4G service signals to deliver them to mobile terminals moving with the speed of 300–350 km/h along the autobahn is given. In this case, the total amount of mobile terminals can be several thousand, and the bandwidth for each signal should be on the order of 6–12 Mbit/s. Two variants of generating a modulating signal without and with using the principles of TDMA technology are offered.

도로 소음 저감을 위한 능동소음제어 시스템의 개발 및 기초실험

PURPOSES : The purpose of this study is about noise which is generated from roads and is consist of irregular frequency variation from low frequency to various band. The existing methods of noise reduction are sound barrier that uses insulation material and absorbing material or have applied passive technology of noise reduction by devices. The total frequency band is needed to apply active noise control. METHODS : In this study applies to the field of road traffic environment, signal processing controller and various analog signal input/output, the amplifier module is based on parallel-core embedded processor designed. DSP performs the control algorithm of the road traffic noise. Noise sources in the open space performance of evaluation were applied. In this study, controller of active signal processor was designed based on the module of audio input/output and main controller of embedded process. The controller of active signal processor operates noise reduction algorithm and performance tests of noise reduction in inside and outside environment were executed. RESULTS : The signal processing controller with OMAP-L137 parallel-core processors as the center, DSP processors in the active control operations dealt with quickly. To maximize the operation speed of an object and ARM processor is external function keys and display for functions and evaluating the performance management system was designed for the purpose of the interface. Therefore the reduction of road traffic noise has established an electronic controller-based noise reduction. CONCLUSIONS : It is shown that noise reduction is effective in the case of pour tonal sound and complex tonal sound below 500Hz by appling to Fx-LMS.

Heart rate variability during the MSLT and MWT in patients with and without excessive daytime sleepiness

Autonomic heart regulation changes during the transition from wakefulness to sleep. Heart rate variability (HRV) can be used to assess the autonomic nervous system function (ANS). We aimed to evaluate ANS in patients with and without excessive daytime sleepiness (EDS) during the maintenance of wakefulness test (MWT) and multiple sleep latency test (MSLT). From a series of 98 consecutive patients with symptoms of sleep disordered breathing (SDB) studied with 5 trials of MWT followed by a research version MSLT two groups of patients (matched by age and gender) were selected consecutively based on mean sleep latencies: the sleepy group (SG) with 17 patients ( MWT < 20 min and MSLT < 8 min) and the alert group (AG) with 20 patients (AG: MWT > 20 and MSLT > 8 min). For each test, the first 120-s window of RR signals during wakefulness was analyzed in the following frequency bands: VLF, < 0.04 Hz; LF, 0.04–0.15 Hz; HF, 0.15–0.4 Hz; TB, total-frequency band. ANS activity was described by measures obtained from traditional power spectral analysis (PSA) and from time–frequency representation (TFR). Non-linear measures – Correntropy (CORR) and auto-mutual-information function (AMIF) – were used to describe the RR regularity. Mean values from all day were compared between groups. Cohen’s effect size was calculated to measure the strength of each difference. Finally, a conditional forward logistic regression analysis was performed to find the best discriminatory variables between groups. TFR (in HF and VLF band), AMIF (in all bands) and CORR (in Total band) analysis showed that the SG had higher values than the AG ( p range 0.000–0.02) during the MSLT. The strongest differences were found in HF band from TRF and AMIF and in Total band from AMIF and CORR. The best sensitivity (82.4%) and specificity (85%) to distinguish patients with and without sleepiness were achieved with the combination of AMIF and CORR in Total band, respectively. No differences were found in traditional PSA measures. Heart rate variability analysis during the first 2 min of the MSLT while the patient is awake show specific changes in TFR and non-linear measures that differentiate sleepy from non-sleepy patients. Study supported by FIS PI07/0318 to MS, cofinanced by FEDER.

Comprehensive Spectral Analysis of HRV during Sleep and their Application in Sleep Staging

Most studies considering spectral features of HRV during sleep divided total frequency band into low frequency (LF, 0.04~0.15Hz) and high frequency (HF, 0.15~0.4 Hz) roughly, and were limited to a few measures like the power in LF and HF, or the ratio of them. To make full use of HRV, more comprehensive spectral features were evaluated in this paper. LF was further divided into true LF (0.04~0.1Hz) and medium frequency (0.1~0.15Hz). Spectrum power, mean frequency and spectral entropy of different spectral bands, fractal dimension and peak in HF (20 measures in total) were calculated for wake, REM, light sleep and deep sleep. The significance between sleep stages of each feature was evaluated. The random forest method was adopted for sleep staging and features importance rank. The results suggested that almost all the new proposed features showed significant differences during different sleep stages. They can improve sleep stages classification performance notably. Our study provided new features for sleep stages classification based on ECG.