Non-stationary Noise Environments Research Articles

Diffusion Augmented Complex Inverse Square Root for Adaptive Frequency Estimation over Distributed Networks

Using adaptive filtering to estimate the frequency of power systems has become a popular trend. In recent years, however, few studies have been performed on adaptive frequency estimations in non-stationary noise environments. In this paper, we propose the distributed complex inverse square root algorithm and distributed augmented complex inverse square root algorithm for the frequency estimation of power systems based on the widely linear model and the inverse square root cost function, where the function can restrain both positive and negative large errors, based on its symmetry. Moreover, the wireless sensor networks support monitoring and adaptation for the frequency estimation in the distributed networks, and the proposed approach can ensure good robustness of the balanced or unbalanced three-phase power system with the help of a local complex-value voltage signal generated by Clark’s transformation. In addition, the bound of step size is driven by the global vectors, and that low computation complexity do not hinder those performances. The results of several experiments demonstrate that our algorithms can effectively estimate the frequency in impulsive noise environments.

Interacting multiple model adaptive robust Kalman filter for process and measurement modeling errors simultaneously

This paper proposes an effective Interactive Multiple Model Adaptive Robust Kalman Filter (IMMARKF) without time delay to handle situations where both process modeling errors and measurement modeling errors exist simultaneously. Building upon the robust Centered Error Entropy Kalman Filter (CEEKF) for outlier measurements and the Adaptive Kalman Filter (AKF) for process modeling errors, the IMMARKF method combines the Gaussian optimality of the KF, the adaptability of AKF, and the robustness of CEEKF using the interacting multiple model (IMM) principle to adapt reasonably to changing application environments, and can obtain estimation results in the absence of time delay. Target tracking simulations show that compared to existing methods, the proposed method can better adapt to non-stationary noise and application environments where process anomalies and measurement anomalies occur simultaneously.

Speech Enhancement Using Sliding Window Empirical Mode Decomposition and Hurst-based Technique

The most challenging in speech enhancement technique is tracking non-stationary noises for long speech segments and low Signal-to-Noise Ratio (SNR). Different speech enhancement techniques have been proposed but, those techniques were inaccurate in tracking highly non-stationary noises. As a result, Empirical Mode Decomposition and Hurst-based (EMDH) approach is proposed to enhance the signals corrupted by non-stationary acoustic noises. Hurst exponent statistics was adopted for identifying and selecting the set of Intrinsic Mode Functions (IMF) that are most affected by the noise components. Moreover, the speech signal was reconstructed by considering the least corrupted IMF. Though it increases SNR, the time and resource consumption were high. Also, it requires a significant improvement under nonstationary noise scenario. Hence, in this article, EMDH approach is enhanced by using Sliding Window (SW) technique. In this SWEMDH approach, the computation of EMD is performed based on the small and sliding window along with the time axis. The sliding window depends on the signal frequency band. The possible discontinuities in IMF between windows are prevented by the total number of modes and the number of sifting iterations that should be set a priori. For each module, the number of sifting iterations is determined by decomposition of many signal windows by standard algorithm and calculating the average number of sifting steps for each module. Based on this approach, the time complexity is reduced significantly with suitable quality of decomposition. Finally, the experimental results show the considerable improvements in speech enhancement under non-stationary noise environments.

Wavelet-Based Weighted Low-Rank Sparse Decomposition Model for Speech Enhancement Using Gammatone Filter Bank Under Low SNR Conditions

Estimating noise-related parameters in unsupervised speech enhancement (SE) techniques is challenging in low SNR and non-stationary noise environments. In the recent SE approaches, the best results are achieved by partitioning noisy speech spectrograms into low-rank noise and sparse speech parts. However, a few limitations reduce the performance of these SE methods due to the use of overlap and add in STFT process, noisy phase, due to inaccurate estimation of low rank in nuclear norm minimization and Euclidian distance measure in the cost function. These aspects can cause a loss of information in the reconstructed signal when compared to clean speech. To solve this, we propose a novel wavelet-based weighted low-rank sparse decomposition model for enhancing speech by incorporating a gammatone filter bank and Kullback–Leibler divergence. The proposed framework differs from other strategies in which the SE is carried entirely in time domain without the need for noise estimation. Further, to reduce the word error rate, these algorithms were trained and tested on a typical automatic speech recognition module. The experimental findings indicate that the proposed cascaded model has shown significant improvement under low SNR conditions over individual and traditional methods with regard to SDR, PESQ, STOI, SIG, BAK and OVL.

Ultra-Short Wave Communication Squelch Algorithm Based on Deep Neural Network

The squelch problem of ultra-short wave communication under non-stationary noise and low Signal-to-Noise Ratio (SNR) in a complex electromagnetic environment is still challenging. To alleviate the problem, we proposed a squelch algorithm for ultra-short wave communication based on a deep neural network and the traditional energy decision method. The proposed algorithm first predicts the speech existence probability using a three-layer Gated Recurrent Unit (GRU) with the speech banding spectrum as the feature. Then it gets the final squelch result by combining the strength of the signal energy and the speech existence probability. Multiple simulations and experiments are done to verify the robustness and effectiveness of the proposed algorithm. We simulate the algorithm in three situations: the typical Amplitude Modulation (AM) and Frequency Modulation (FM) in the ultra-short wave communication under different SNR environments, the non-stationary burst-like noise environments, and the real received signal of the ultra-short wave radio. The experimental results show that the proposed algorithm performs better than the traditional squelch methods in all the simulations and experiments. In particular, the false alarm rate of the proposed squelch algorithm for non-stationary burst-like noise is significantly lower than that of traditional squelch methods.

Multi-target ensemble learning based speech enhancement with temporal-spectral structured target

Recently, deep neural network (DNN)-based speech enhancement has shown considerable success, and mapping-based and masking-based are the two most commonly used methods. However, these methods do not consider the spectrum structures of signal. In this paper, a novel structured multi-target ensemble learning (SMTEL) framework is proposed, which uses target temporal-spectral structures to improve speech quality and intelligibility. First, the basis matrices of clean speech, noise, and ideal ratio mask (IRM) are captured by the sparse nonnegative matrix factorization, which contain the basic structures of the signal. Second, the basis matrices are co-trained with a multi-target DNN to estimate the activation matrices instead of directly estimating the targets. Then a joint training single layer perceptron is proposed to integrate the two targets and further improve speech quality and intelligibility. The sequential floating forward selection method is used to systematically analyze the impact of the integrated targets on enhanced performance, and analyze the effect of the target weights on the results. Finally, the proposed method with progressive learning is combined to improve the enhanced performance. Systematic experiments on the UW/NU corpus show that the proposed method achieves the best enhancement effect in the case of low network cost and complexity, especially in visible nonstationary noise environment. Compared with the target integration method which does not use structured targets and the long short-term memory masking method, the speech quality of the proposed method is improved by 25.6 % and 29.2 % of restaurant noise, and the speech intelligibility is improved by 35.5 % and 15.8 %, respectively.

Genetic Algorithm-Based Adaptive Wiener Gain for Speech Enhancement Using an Iterative Posterior NMF

In this paper, we propose a genetic algorithm-based adaptive Wiener gain for speech enhancement using an iterative posterior non-negative matrix factorization (NMF). In the recent past, NMF-based Wiener filtering methods were used to improve the performance of speech enhancement, which has shown that they provide better performance when compared with conventional NMF methods. But performance degrades in non-stationary noise environments. Template-based approaches are more robust and perform better in non-stationary noise environments compared to statistical model-based approaches but are dependent on a priori information. Combining the approaches avoids the drawbacks of both. To improve the performance further, speech and noise bases are adapted simultaneously in the NMF approach. The usage of Super-Gaussian constraints in iterative NMF still improves the performance in non-stationary noise. The silence frame is a challenging task in the case of NMF; still there will be some amount of noise present in those frames. For further enhancement, we have combined with a genetic algorithm (GA)-based adaptive Wiener filter which performs well in denoising and also the GA search the adaptive [Formula: see text] allows us to control the trade-off between fitting the observed spectrogram of mixed speech and noise achieving high likelihood under our prior model. The proposed method outperforms other benchmark algorithms in terms of the source to distortion ratio (SDR), short-time objective intelligibility (STOI), and perceptual evaluation of speech quality (PESQ).

Generalized Subspace Snoring Signal Enhancement Based on Noise Covariance Matrix Estimation

Acoustical properties of snoring signal have been widely studied as a potentially cost-effective and reliable alternative to diagnosing obstructive sleep apnea hypopnea syndrome, with a common recognition that the diagnostic accuracy depends heavily on the snoring signal quality. In the paper, generalized subspace noise reduction based on noise covariance matrix estimate is proposed. The noise covariance matrix is the Toeplitz matrix of the unbiased autocorrelation sequence which is estimated by recursive averaging its past value adjusted by a time-varying smoothing parameter controlled by the snoring signal presence probability, and the signal presence is determined by the ratio of temporal frame autocorrelation value to its minimum absolute value. The proposed method has a better estimate of noise covariance matrix, and the results of objective quality measurements and spectrograms of snoring signal show obvious improvement in terms of noise reduction and signal distortion under different non-stationary noise environments compared with conventional subspace enhancement algorithm.

Automated two-dimensional localization of underwater acoustic transient impulses using vector sensor image processing (vector sensor localization).

Detecting acoustic transients by signal-to-noise ratio (SNR) becomes problematic in nonstationary ambient noise environments characteristic of coral reefs. An alternate approach presented here uses signal directionality to automatically detect and localize transient impulsive sounds collected on underwater vector sensors spaced tens of meters apart. The procedure, which does not require precise time synchronization, first constructs time-frequency representations of both the squared acoustic pressure (spectrogram) and dominant directionality of the active intensity (azigram) on each sensor. Within each azigram, sets of time-frequency cells associated with transient energy arriving from a consistent azimuthal sector are identified. Binary image processing techniques then link sets that share similar duration and bandwidth between different sensors, after which the algorithm triangulates the source location. Unlike most passive acoustic detectors, the threshold criterion for this algorithm is bandwidth instead of pressure magnitude. Data collected from shallow coral reef environments demonstrate the algorithm's ability to detect SCUBA bubble plumes and consistent spatial distributions of somniferous fish activity. Analytical estimates and direct evaluations both yield false transient localization rates from 3% to 6% in a coral reef environment. The SNR distribution of localized pulses off Hawaii has a median of 7.7 dB and interquartile range of 7.1 dB.

A Real-Time Speech Separation Method Based on Camera and Microphone Array Sensors Fusion Approach.

In the context of assisted human, identifying and enhancing non-stationary speech targets speech in various noise environments, such as a cocktail party, is an important issue for real-time speech separation. Previous studies mostly used microphone signal processing to perform target speech separation and analysis, such as feature recognition through a large amount of training data and supervised machine learning. The method was suitable for stationary noise suppression, but relatively limited for non-stationary noise and difficult to meet the real-time processing requirement. In this study, we propose a real-time speech separation method based on an approach that combines an optical camera and a microphone array. The method was divided into two stages. Stage 1 used computer vision technology with the camera to detect and identify interest targets and evaluate source angles and distance. Stage 2 used beamforming technology with microphone array to enhance and separate the target speech sound. The asynchronous update function was utilized to integrate the beamforming control and speech processing to reduce the effect of the processing delay. The experimental results show that the noise reduction in various stationary and non-stationary noise environments were 6.1 dB and 5.2 dB respectively. The response time of speech processing was less than 10ms, which meets the requirements of a real-time system. The proposed method has high potential to be applied in auxiliary listening systems or machine language processing like intelligent personal assistant.

Speech enhancement algorithm of improved OMLSA based on bilateral spectrogram filtering

In this paper, a bilateral spectrogram filtering (BSF)-based optimally modified log-spectral amplitude (OMLSA) estimator for single-channel speech enhancement is proposed, which can significantly improve the performance of OMLSA, especially in highly non-stationary noise environments, by taking advantage of bilateral filtering (BF), a widely used technology in image and visual processing, to preprocess the spectrogram of the noisy speech. BSF is capable of not only sharpening details, removing unwanted textures or background noise from the noisy speech spectrogram, but also preserving edges when considering a speech spectrogram as an image. The a posteriori signal-to-noise ratio (SNR) of OMLSA algorithm is estimated after applying BSF to the noisy speech. Besides, in order to reduce computing costs, a fast and accurate BF is adopted to reduce the algorithm complexity O(1) for each time-frequency bin. Finally, the proposed algorithm is compared with the original OMLSA and other classic denoising methods using various types of noise with different signal-to-noise ratios in terms of objective evaluation metrics such as segmental signal-to-noise ratio improvement and perceptual evaluation of speech quality. The results show the validity of the improved BSF-based OMLSA algorithm.

DOA estimation for uncorrelated and coherent signals in non-stationary noise environments

ABSTRACTIn this paper, a new method, namely double spatial covariance difference (DSCD), is presented for DOA estimation of uncorrelated and coherent signals in time varying noise environments. In the proposed method, the effect of the noise is eliminated and therefore DSCD can be used in general noise, as well as low SNR environments. In addition, computational burden has been reduced regarding to no use of parameter and peak search or iterative calculation. Simulation results demonstrate the performance of the proposed method.

Multi-resolution auditory cepstral coefficient and adaptive mask for speech enhancement with deep neural network

The performance of the existing speech enhancement algorithms is not ideal in low signal-to-noise ratio (SNR) non-stationary noise environments. In order to resolve this problem, a novel speech enhancement algorithm based on multi-feature and adaptive mask with deep learning is presented in this paper. First, we construct a new feature called multi-resolution auditory cepstral coefficient (MRACC). This feature which is extracted from four cochleagrams of different resolutions can capture the local information and spectrotemporal context and reduce the algorithm complexity. Second, an adaptive mask (AM) which can track noise change for speech enhancement is put forward. The AM can flexibly combine the advantages of an ideal binary mask (IBM) and an ideal ratio mask (IRM) with the change of SNR. Third, a deep neural network (DNN) architecture is used as a nonlinear function to estimate adaptive mask. And the first and second derivatives of MRACC and MRACC are used as the input of the DNN. Finally, the estimated AM is used to weight the noisy speech to achieve enhanced speech. Experimental results show that the proposed algorithm not only further improves speech quality and intelligibility, but also suppresses more noise than the contrast algorithms. In addition, the proposed algorithm has a lower complexity than the contrast algorithms.

A novel fast nonstationary noise tracking approach based on MMSE spectral power estimator

Estimating the noise power spectral density (PSD) from the corrupted speech signal is an essential component for speech enhancement algorithms. In this paper, a novel noise PSD estimation algorithm based on minimum mean-square error (MMSE) is proposed. The noise PSD estimate is obtained by recursively smoothing the MMSE estimation of the current noise spectral power. For the noise spectral power estimation, a spectral weighting function is derived, which depends on the a priori signal-to-noise ratio (SNR). Since the speech spectral power is highly important for the a priori SNR estimate, this paper proposes an MMSE spectral power estimator incorporating speech presence uncertainty (SPU) for speech spectral power estimate to improve the a priori SNR estimate. Moreover, a bias correction factor is derived for speech spectral power estimation bias. Then, the estimated speech spectral power is used in “decision-directed” (DD) estimator of the a priori SNR to achieve fast noise tracking. Compared to three state-of-the-art approaches, i.e., minimum statistics (MS), MMSE-based approach, and speech presence probability (SPP)-based approach, it is clear from experimental results that the proposed algorithm exhibits more excellent noise tracking capability under various nonstationary noise environments and SNR conditions. When employed in a speech enhancement system, improved speech enhancement performances in terms of segmental SNR improvements (SSNR+) and perceptual evaluation of speech quality (PESQ) can be observed.

Personal Identification Using a Robust Eigen ECG Network Based on Time-Frequency Representations of ECG Signals

This paper is concerned with personal identification using a robust EigenECG network (REECGNet) based on time-frequency representations of electrocardiogram (ECG) signals. For this purpose, we use a robust principal component analysis network (RPCANet) and wavelet analysis. In general, PCA performance and applicability in real case scenarios is limited by the lack of robustness to outliers and corrupted observations. However, in a real nonstationary ECG noise environment, RPCA shows good performance when the method is applied with variable dimensions of local signal subspaces. That is why RPCA-based ECG identification is extremely robust with nonlinear data. Also, a REECGNet performs well without back-propagation to obtain features from the visual content. We constructed a Chosun University ECG Database (CU-ECG DB) and compared with the Physikalisch-Technische Bundesanstalt ECG database (PTB-ECG DB), which is shared data. Finally, the experimental results show the advantages and effectiveness of the applied recognition scheme with 98.25% performance. In addition, to demonstrate the superiority of REECGNet, we experimented with adding noise and the experimental result showed 97.5% recognition rate.

Continuous Tamil Speech Recognition technique under non stationary noisy environments

In the last few years, the need for Continuous Speech Recognition system in Tamil language has been increased widely. In this research work, efficient Continuous Tamil Speech Recognition (CTSR) technique is proposed under non stationary noisy environments. This research work consists of two stages such as speech enhancement and modelling phase. In this, the modified Modulation Magnitude Estimation based Spectral Subtraction with Chi-Square Distribution based Noise Estimation (SS–NE) algorithm is proposed to enhance the noisy Tamil speech signal under various non-stationary noise environments. In order to extract the speech segments from the continuous speech, further the enhanced speech signal is segmented through the combination of short-time signal energy and spectral centroid features of the signal. In this work, 26 mel frequency cepstral coefficients per frame are found as optimal values and they are considered as acoustic feature vectors for each frame. In this research work, the Fuzzy C-Means (FCM) clustering is used in order to cluster the extracted feature vectors into discrete symbols. From the evaluation results, it is found that the optimal number of clusters ‘C’ as 5. Finally, Tamil speech from various speakers is recognized using Expectation Maximization Gaussian Mixture Model (EM-GMM) with 16 component densities under continuous measurements of labelled features from FCM clustering techniques in order to reduce the word error rate. From the simulated results, it is observed that the proposed FCM with EM-GMM model for CTSR improves the recognition accuracy from 1.2 to 4.4% when compared to the existing algorithms under different noisy environments by reducing the WER from 1.6 to 5.47%.

Laplacian Speech Model and Soft Decision Based MMSE Estimator for Noise Power Spectral Density in Speech Enhancement

The estimation of noise Power spectral density (PSD) is a very crucial issue for speech enhancement as a result of its significant effiect on the quality and intelligibility of the enhanced speech. Most of the existing estimators for noise PSD try to employ Gaussian speech priors, which, however, have been proven inconsistent with the reality. We derived an effiective solution to this problem of estimating noise PSD in the Minimum mean square error (MMSE) sense when the speech component is modeled by a Laplacian distribution. Meanwhile, the soft decision technique instead of the hard Voice activity detection (VAD) is evolved into our algorithm, which can automatically makes the estimation unbiased without requiring a bias compensation. The performance of the proposed method is tested by several objective and subjective measures under various stationary and nonstationary noise environments. The results confirm that our method achieves good performance for all the noise conditions and Signalnoise-ratio (SNR) settings.

Evidence-Based Occupational Hearing Screening I: Modeling the Effects of Real-World Noise Environments on the Likelihood of Effective Speech Communication.

The objectives of this study were to (1) identify essential hearing-critical job tasks for public safety and law enforcement personnel; (2) determine the locations and real-world noise environments where these tasks are performed; (3) characterize each noise environment in terms of its impact on the likelihood of effective speech communication, considering the effects of different levels of vocal effort, communication distances, and repetition; and (4) use this characterization to define an objective normative reference for evaluating the ability of individuals to perform essential hearing-critical job tasks in noisy real-world environments. Data from five occupational hearing studies performed over a 17-year period for various public safety agencies were analyzed. In each study, job task analyses by job content experts identified essential hearing-critical tasks and the real-world noise environments where these tasks are performed. These environments were visited, and calibrated recordings of each noise environment were made. The extended speech intelligibility index (ESII) was calculated for each 4-sec interval in each recording. These data, together with the estimated ESII value required for effective speech communication by individuals with normal hearing, allowed the likelihood of effective speech communication in each noise environment for different levels of vocal effort and communication distances to be determined. These likelihoods provide an objective norm-referenced and standardized means of characterizing the predicted impact of real-world noise on the ability to perform essential hearing-critical tasks. A total of 16 noise environments for law enforcement personnel and eight noise environments for corrections personnel were analyzed. Effective speech communication was essential to hearing-critical tasks performed in these environments. Average noise levels, ranged from approximately 70 to 87 dBA in law enforcement environments and 64 to 80 dBA in corrections environments. The likelihood of effective speech communication at communication distances of 0.5 and 1 m was often less than 0.50 for normal vocal effort. Likelihood values often increased to 0.80 or more when raised or loud vocal effort was used. Effective speech communication at and beyond 5 m was often unlikely, regardless of vocal effort. ESII modeling of nonstationary real-world noise environments may prove an objective means of characterizing their impact on the likelihood of effective speech communication. The normative reference provided by these measures predicts the extent to which hearing impairments that increase the ESII value required for effective speech communication also decrease the likelihood of effective speech communication. These predictions may provide an objective evidence-based link between the essential hearing-critical job task requirements of public safety and law enforcement personnel and ESII-based hearing assessment of individuals who seek to perform these jobs.

Evidence-based occupational hearing screening II: validation of a screening methodology using measures of functional hearing ability

Objective: Validate use of the Extended Speech Intelligibility Index (ESII) for prediction of speech intelligibility in non-stationary real-world noise environments. Define a means of using these predictions for objective occupational hearing screening for hearing-critical public safety and law enforcement jobs. Design: Analyses of predicted and measured speech intelligibility in recordings of real-world noise environments were performed in two studies using speech recognition thresholds (SRTs) and intelligibility measures. ESII analyses of the recordings were used to predict intelligibility. Noise recordings were made in prison environments and at US Army facilities for training ground and airborne forces. Speech materials included full bandwidth sentences and bandpass filtered sentences that simulated radio transmissions. Study sample: A total of 22 adults with normal hearing (NH) and 15 with mild–moderate hearing impairment (HI) participated in the two studies. Results: Average intelligibility predictions for individual NH and HI subjects were accurate in both studies (r2 ≥ 0.94). Pooled predictions were slightly less accurate (0.78 ≤ r2 ≤ 0.92). Conclusions: An individual’s SRT and audiogram can accurately predict the likelihood of effective speech communication in noise environments with known ESII characteristics, where essential hearing-critical tasks are performed. These predictions provide an objective means of occupational hearing screening.

Pitch-Based Voice Activity Detection for Feedback Cancellation and Noise Reduction in Hearing Aids

In hearing aid (HA) systems, amplification of speech signals is done to compensate the hearing loss of patients. Background noise and feedback signals may also get amplified which degrade the intelligibility and quality of speech. To achieve high de-noise efficiency, signal processing unit in HA system has voice activity detector (VAD). The conventional VAD detects voice based on zero crossing rate or energy of input signals. However, these methods cannot perform well at low SNR or non-stationary noise environments. Since pitch is a special characteristic of speech and is basically independent of noise intensity, VAD based on pitch can have high accuracy even when the noise spectrum is changing drastically. In this paper, pitch-based VAD is presented and its accuracy is checked against zero crossing rate-based VAD (ZCR-VAD). For noise reduction, an improved multi-band spectral over-subtraction algorithm is employed along with the high accurate pitch-based VAD. For feedback cancellation, the performance of adaptive algorithms like NLMS, RLS and affine projection (AP) algorithms with pitch-based VAD is compared and it is observed that AP is suitable for feedback cancellation. The proposed noise reduction and feedback cancellation algorithm with pitch-based VAD method is tested with NOIZEUS speech database and with real-time noisy speech signals. The simulation results show that the accuracy of pitch-VAD is about 23% higher than that of ZCR-VAD. The SNR for the proposed noise reduction and feedback cancellation with pitch-VAD method is improved by 10 dB than conventional spectral subtraction and adaptive algorithms. Mean opinion score (MOS) obtained for the proposed method is 4.3 out of 5.