Signal-to-distortion Ratio Research Articles

End-to-end underwater acoustic source separation model based on EDBG-GALR

In practical work scenarios, it is common for multiple vessels to be present in close proximity within the same water area. In such cases, the acoustic signals emitted by these vessels often overlap, causing interference and reducing the accuracy of vessel identification. This paper proposes an improved GALR end-to-end underwater acoustic source separation algorithm, with ECA-DE (Efficient Channel Attention-Deep Encoder) and Bi-GRU (Bidirectional Gated Recurrent Unit), which consists of an encoder, separation blocks, and a decoder (EDBG-GALR). Addressing the issue of GALR encoder’s limited expressiveness due to its single-layer one-dimensional convolutional layer, we introduce a new deep encoder, ECA-DE, to enhance the encoder’s capability to process temporal signals and improve the efficiency of the separator’s input. Additionally, we integrate Bi-GRU into the GALR separation block’s local modeling module to enhance the local modeling ability of sequence features, thereby reducing the model’s computational and parameter requirements. Experimental results demonstrate that the proposed EDBG-GALR method can effectively separate mixed multi-target signals into multiple single-target signals, achieving a maximum scale-invariant signal-to-noise ratio (SI-SNR) improvement of 3.32 dB and a signal distortion ratio (SDR) improvement of 3.38 dB over baseline methods. These results highlight the practical applicability of EDBG-GALR in complex underwater environments.

A 68.5 dB-SNDR, 12.4-fJ/conv.-step, 100-MS/s pipelined-SAR ADC with PVT-enhanced circuitry

This paper presents a signal-chain friendly 12-bit pipelined-successive approximation register (SAR) analog-to-digital converter (ADC) in 40nm CMOS, which is optimized to achieve a 68.5 dB-SNDR, 100-MS/s sample rate with 54.5% of maximum available input range. The implemented ADC employs an improved clock booster and latch-register to achieve high on-conductance of nmos switches and low-leakage data storage. It also explores an adaptive non-overlapping complementary clock generator and a dedicated VCM buffer to accommodate process, voltage, and temperature (PVT) conditions. The relative variation of non-overlapping time is reduced by 50% compared to the conventional method, and the signal-to-distortion ratio (SDR) of residue amplification is improved by 8dB . Moreover, on-chip bit weight calibration is implemented to address the gain error brought by capacitor mismatch and inner-stage gain error. The prototype ADC is simulated under 5 corners, −40 to 125°C, and 1.1V±2.5%. For a Nyquist input, the simulated SNDR under typical corner is 68.5dB , which can remain over 66dB under PVT conditions. The achieved Walden Figure of Merit (FoM) is 12.4-fJ/conv.-step.

A bidirectional domain separation adversarial network based transfer learning method for near-infrared spectra

Traditional calibration transfer (CT) methods usually fail to adapt the source domain model to the target domain because of changes associated with the instrument, detection environment, sample composition, or sample type. Most deep transfer learning (DTL) methods have shown reliable capability in extracting domain-invariant representations. However, factors unique to the spectral domain limit their development and application, including difficulty in obtaining labeled samples, numerous spectral perturbation factors, and the scarcity of spectral databases. To address these challenges, we here introduce the Bidirectional Domain-Separating Adversarial Network (BiDSAN), based on a deep adversarial network transfer architecture, for learning and adapting to distribution/domain differences between sample spectra. BiDSAN incorporates Wasserstein divergence (w-div) and dynamic adversarial factor into the domain critic (discriminator) to accelerate the extraction of spectral shared components. Additionally, the use of the spectral signal distortion ratio as a loss function and the introduction of a bidirectional mechanism further enhanced the training accuracy and speed of spectral models. Calibration analyses were tested on pharmaceutical and two sets of wood datasets, corresponding to three common domain/distribution difference scenarios discussed in the study. Compared to existing CT and DTL methods, the proposed BiDSAN method demonstrated superior spectral signal calibration capability and stability, achieving unsupervised domain adaptation for similar samples and, for the first time, obtaining ideal prediction results in cross-species spectral domain adaptation. Our method represents significant potential for using DTL approaches in spectral domain adaptation.

Iteratively Refined Multi-Channel Speech Separation

The combination of neural networks and beamforming has proven very effective in multi-channel speech separation, but its performance faces a challenge in complex environments. In this paper, an iteratively refined multi-channel speech separation method is proposed to meet this challenge. The proposed method is composed of initial separation and iterative separation. In the initial separation, a time–frequency domain dual-path recurrent neural network (TFDPRNN), minimum variance distortionless response (MVDR) beamformer, and post-separation are cascaded to obtain the first additional input in the iterative separation process. In iterative separation, the MVDR beamformer and post-separation are iteratively used, where the output of the MVDR beamformer is used as an additional input to the post-separation network and the final output comes from the post-separation module. This iteration of the beamformer and post-separation is fully employed for promoting their optimization, which ultimately improves the overall performance. Experiments on the spatialized version of the WSJ0-2mix corpus showed that our proposed method achieved a signal-to-distortion ratio (SDR) improvement of 24.17 dB, which was significantly better than the current popular methods. In addition, the method also achieved an SDR of 20.2 dB on joint separation and dereverberation tasks. These results indicate our method’s effectiveness and significance in the multi-channel speech separation field.

Improved recovery of cardiac auscultation sounds using modified cosine transform and LSTM-based masking.

Obtaining accurate cardiac auscultation signals, including basic heart sounds (S1 and S2) and subtle signs of disease, is crucial for improving cardiac diagnoses and making the most of telehealth. This research paper introduces an innovative approach that utilizes a modified cosine transform (MCT) and a masking strategy based on long short-term memory (LSTM) to effectively distinguish heart sounds and murmurs from background noise and interfering sounds. The MCT is used to capture the repeated pattern of the heart sounds, while the LSTMs are trained to construct masking based on the repeated MCT spectrum. The proposed strategy's performance in maintaining the clinical relevance of heart sounds continues to demonstrate effectiveness, even in environments marked by increased noise and complex disruptions. The present work highlights the clinical significance and reliability of the suggested methodology through in-depth signal visualization and rigorous statistical performance evaluations. In comparative assessments, the proposed approach has demonstrated superior performance compared to recent algorithms, such as LU-Net and PC-DAE. Furthermore, the system's adaptability to various datasets enhances its reliability and practicality. The suggested method is a potential way to improve the accuracy of cardiovascular diagnostics in an era of rapid advancement in medical signal processing. The proposed approach showed an enhancement in the average signal-to-noise ratio (SNR) by 9.6dB at an input SNR of - 6dB and by 3.3dB at an input SNR of 10dB. The average signal distortion ratio (SDR) achieved across a variety of input SNR values was 8.56dB.

An In-depth Review on Music Source Separation

The study in Music Source Separation (MSS) raises a fundamental question: Is there any benefit in considering broader contextual information, or are local acoustic features adequate? In various domains, attention-based Transformers [1] have demonstrated their capacity to assimilate information across extensive sequences. In our research, we introduce Hybrid Transformer Demucs (HT Demucs), a hybrid temporal/spectral bi-U-Net based on Hybrid Demucs [2]. Here, the innermost layers are substituted with a cross-domain Transformer Encoder, utilizing self-attention within one domain and cross-attention across domains. Although its performance is lacking when exclusively trained on MUSDB [3], we illustrate that it surpasses Hybrid Demucs (trained on the same data) by 0.45 dB of Signal-to-Distortion Ratio (SDR) when provided with an additional 800 training songs. By employing sparse attention kernels to broaden its receptive field and undertaking per-source fine-tuning, we attain state-of-the-art results on MUSDB with extra training data, achieving a remarkable 9.20 dB of SDR.

Hidden Follower Detection: How Is the Gaze-Spacing Pattern Embodied in Frequency Domain?

Spatiotemporal social behavior analysis is a technique that studies the social behavior patterns of objects and estimates their risks based on their trajectories. In social public scenarios such as train stations, hidden following behavior has become one of the most challenging issues due to its probability of evolving into violent events, which is more than 25%. In recent years, research on hidden following detection (HFD) has focused on differences in time series between hidden followers and normal pedestrians under two temporal characteristics: gaze and spatial distance. However, the time-domain representation for time series is irreversible and usually causes the loss of critical information. In this paper, we deeply study the expression efficiency of time/frequency domain features of time series, by exploring the recovery mechanism of features to source time series, we establish a fidelity estimation method for feature expression and a selection model for frequency-domain features based on the signal-to-distortion ratio (SDR). Experimental results demonstrate the feature fidelity of time series and HFD performance are positively correlated, and the fidelity of frequency-domain features and HFD performance are significantly better than the time-domain features. On both real and simulated datasets, the accuracy of the proposed method is increased by 3%, and the gaze-only module is improved by 10%. Related research has explored new methods for optimal feature selection based on fidelity, new patterns for efficient feature expression of hidden following behavior, and the mechanism of multimodal collaborative identification.

On intrusive speech quality measures and a global SNR based metric

Measuring the quality of noisy speech signals has been an increasingly important problem in the field of speech processing as more and more speech-communication and human-machine-interface systems are deployed in practical applications. In this paper, we study four widely used classical performance measures: signal-to-distortion ratio (SDR), short-time objective intelligibility (STOI), signal-to-noise ratio (SNR), and perceptual evaluation of speech quality (PESQ). Through analyzing these performance measures under the same framework and identifying the relationship between their core parameters, we convert these measures into the corresponding equivalent SNRs. This conversion enables not only some new insights into different quality measures but also a way to combine these measures into a new metric. In the derivation of the equivalent SNRs, we introduce the widely used masking technique into the computation of correlation coefficients, which is subsequently used to analyze STOI. Furthermore, we propose an attention method to compute the core parameters of PESQ, and also an empirical formula to project the equivalent SNRs into PESQ scores. Experiments are carried out and the results justifies the properties of the derived quality measures.

Analysis of statistical estimators and neural network approaches for speech enhancement

Speech communicated is adversely affected by environmental noise. It is important to process the speech and reduce noise for better understanding. Speech enhancement or noise reduction is useful to provide comfort for human or machine listening. Traditional algorithms provide better noise reduction and better-quality speech. Due to the non-stationary nature of noise and the quasi-stationary nature of speech, the traditional methods are proven inadequate in achieving high-quality speech. Later statistical estimators based on Gaussian, and super-Gaussian Probability Density Function (PDF) assumption further improved the enhancement performance. But still, non-stationary noise nature introduces artifacts in processed signal and results in decreased performance. It is observed that neural network approaches and the factorization approach provide better performance even under non-stationary noises by proper training and large database. Different features result in variations in output performance under unseen noise and speaker conditions. It is important to understand the importance and advantages of traditional methods, statistical estimators, and neural network approaches performances. To select the suitable method for a required application, it is essential to consider the trade-off between quality and distortion. In this work, the importance of speech enhancement methods is discussed. Performance measures used for understanding the speech enhancement like Signal to Noise Ratio (SNR), Segmental SNR, Log-Likelihood Ratio (LLR), Weighted Spectral Slope (WSS), Perceptual Evaluation of Speech Quality (PESQ), Short-Time Objective Intelligibility (STOI), Signal to Distortion Ratio (SDR) and Mean Opinion Score (MOS), are given. Highlights of important results are discussed for analyzing better speech enhancement methods for the required application. In this work, performance is compared using objective and subjective performance measures. Simulation results show superior performance when neural network is employed in statistical estimators.

Sparse Signal Recovery through Long Short-Term Memory Networks for Compressive Sensing-Based Speech Enhancement

This paper presents a novel speech enhancement approach based on compressive sensing (CS) which uses long short-term memory (LSTM) networks for the simultaneous recovery and enhancement of the compressed speech signals. The advantage of this algorithm is that it does not require an iterative process to recover the compressed signals, which makes the recovery process fast and straight forward. Furthermore, the proposed approach does not require prior knowledge of signal and noise statistical properties for sensing matrix optimization because the used LSTM can directly extract and learn the required information from the training data. The proposed technique is evaluated against white, babble, and f-16 noises. To validate the effectiveness of the proposed approach, perceptual evaluation of speech quality (PESQ), short-time objective intelligibility (STOI), and signal-to-distortion ratio (SDR) were compared to other variants of OMP-based CS algorithms The experimental outcomes show that the proposed approach achieves the maximum improvements of 50.06%, 43.65%, and 374.16% for PESQ, STOI, and SDR respectively, over the different variants of OMP-based CS algorithms.

A comparative study of multiple music signal noise reduction algorithms in music processing

Electronic music is susceptible to noise in production, which needs to be processed. This paper analyzed several commonly used noise reduction algorithms, including wiener filtering, wavelet transform, spectral subtraction, and improved spectral subtraction, and then compared the noise reduction performance of several algorithms by producing noisy music datasets in the audio analysis tool librosa. It was found from the experimental results that the wavelet transform algorithm performed best when sym3 was used as the wavelet basis function, and the number of decomposition layers was 7. The comparison of different algorithms showed that the performance of the wiener filtering algorithm was poor in reducing noise, and the signal-to-noise ratio (SNR) and signal distortion ratio (SDR) was low; the performance of the improved spectral subtraction algorithm was the best, and the SNR and SDR were 20.36 dB and 17.94 dB, respectively, when the SNR was −8 dB, which were better than the other algorithms. The experimental results demonstrate the reliability of the improved spectral subtraction method in music signal noise reduction. The algorithm can be applied in practical music processing.

Single-channel speech separation using soft-minimum permutation invariant training

The goal of speech separation is to extract multiple speech sources from a single microphone recording. Recently, with the advancement of deep learning and availability of large datasets, speech separation has been formulated as a supervised learning problem. These approaches aim to learn discriminative patterns of speech, speakers, and background noise using a supervised learning algorithm, typically a deep neural network. A long-lasting problem in supervised speech separation is finding the correct label for each separated speech signal, referred to as label permutation ambiguity. Permutation ambiguity refers to the problem of determining the output-label assignment between the separated sources and the available single-speaker speech labels. Finding the best output-label assignment is required for calculation of separation error, which is later used for updating parameters of the model. Recently, Permutation Invariant Training (PIT) has been shown to be a promising solution in handling the label ambiguity problem. However, the overconfident choice of the output-label assignment by PIT results in a sub-optimal trained model. In this work, we propose a probabilistic optimization framework to address the inefficiency of PIT in finding the best output-label assignment. Our proposed method entitled trainable Soft-minimum PIT is then employed on the same Long-Short Term Memory (LSTM) architecture used in Permutation Invariant Training (PIT) speech separation method. The results of our experiments show that the proposed method outperforms conventional PIT speech separation significantly (p-value <0.01) by +1dB in Signal to Distortion Ratio (SDR) and +1.5dB in Signal to Interference Ratio (SIR).

Alpha-divergence two-dimensional nonnegative matrix factorization for biomedical blind source separation

An alpha-divergence two-dimensional nonnegative matrix factorization (NMF2D) for biomedical signal separation is presented. NMF2D is a popular approach for retrieving low-rank approximations of nonnegative data such as image pixel, audio signal, data mining, pattern recognition and so on. In this paper, we concentrate on biomedical signal separation by using NMF2D with alpha-divergence family which decomposes a mixture into two-dimensional convolution factor matrices that represent temporal code and the spectral basis. The proposed iterative estimation algorithm (alpha-divergence algorithm) is initialized with random values, and it updated using multiplicative update rules until the values converge. Simulation experiments were carried out by comparing the original and estimated signal in term of signal-to-distortion ratio (SDR). The performances have been evaluated by including and excluding the sparseness constraint which sparseness is favored by penalizing nonzero gains. As a result, the proposed algorithm improved the iteration speed and sparseness constraints produce slight improvement of SDR.

Monaural cardiopulmonary sound separation via complex-valued deep autoencoder and cyclostationarity

Objective. Cardiopulmonary auscultation is promising to get smart due to the emerging of electronic stethoscopes. Cardiac and lung sounds often appear mixed at both time and frequency domain, hence deteriorating the auscultation quality and the further diagnosis performance. The conventional cardiopulmonary sound separation methods may be challenged by the diversity in cardiac/lung sounds. In this study, the data-driven feature learning advantage of deep autoencoder and the common quasi-cyclostationarity characteristic are exploited for monaural separation. Approach. Different from most of the existing separation methods that only handle the amplitude of short-time Fourier transform (STFT) spectrum, a complex-valued U-net (CUnet) with deep autoencoder structure, is built to fully exploit both the amplitude and phase information. As a common characteristic of cardiopulmonary sounds, quasi-cyclostationarity of cardiac sound is involved in the loss function for training. Main results. In experiments to separate cardiac/lung sounds for heart valve disorder auscultation, the averaged achieved signal distortion ratio (SDR), signal interference ratio (SIR), and signal artifact ratio (SAR) in cardiac sounds are 7.84 dB, 21.72 dB, and 8.06 dB, respectively. The detection accuracy of aortic stenosis can be raised from 92.21% to 97.90%. Significance. The proposed method can promote the cardiopulmonary sound separation performance, and may improve the detection accuracy for cardiopulmonary diseases.

Noise Suppression Using Beamformer and Transfer-Function-Gain Nonnegative Matrix Factorization with Distributed Stereo Microphones

In this paper, we propose a novel approach to noise suppression using multiple distributed recording devices with stereo microphones. In the proposed method, noise suppression based on phase information is applied to the synchronous stereo signals captured by each recording device and then output signals are utilized for transfer-function-gain nonnegative matrix factorization (NMF) as extra input signals. We intended to estimate the target signal more accurately by transfer-function-gain NMF. Experiments using impulse responses measured in a meeting room have shown that the proposed method outperformed conventional methods using transfer-function-gain NMF in terms of the signal-to-distortion ratio (SDR) and signal-to-interference ratio (SIR).

Single channel blind source separation of complex signals based on spatial‐temporal fusion deep learning

Blind Source Separation (BSS) of complex signals composed of radar, communication and jamming signals is the first step in an integrated electronic system, which requires higher accuracy of separation. However, the traditional Single-Channel Blind Source Separation (SCBSS) method has low separation accuracy and poor robustness. Aiming at this problem, this paper proposes a SCBSS method based on spatial-temporal fusion deep learning model. This is a deep neural network model, which realizes spatial-temporal of mixed signals by integrating Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory Network (BiLSTM). Convolutional Neural Network is used to extract spatial features from input sequences, and BiLSTM is used to mine timing rules of signals. A batch normalisation layer and a dropout layer are added to improve stability and prevent overfitting. The experiments show that the average similarity coefficient of the separated signals is above 0.99 and the Signal-Distortion Ratio (SDR) is up to 27 dB without noise. When the Signal-Noise Ratio is 0–20 dB and Jamming-Signal Ratio is 15 dB, the SDR is 5–30 dB higher than the traditional methods and the single network structure deep learning methods.

Single-Channel Speech Separation Using Soft-Minimum Permutation Invariant Training

The goal of speech separation is to extract multiple speech sources from a single microphone recording. Recently, with the advancement of deep learning and availability of large datasets, speech separation has been formulated as a supervised learning problem. These approaches aim to learn discriminative patterns of speech, speakers, and background noise using a supervised learning algorithm, typically a deep neural network. A long-lasting problem in supervised speech separation is finding the correct label for each separated speech signal, referred to as label permutation ambiguity. Permutation ambiguity refers to the problem of determining the output-label assignment between the separated sources and the available single-speaker speech labels. Finding the best output-label assignment is required for calculation of separation error, which is later used for updating parameters of the model. Recently, Permutation Invariant Training (PIT) has been shown to be a promising solution in handling the label ambiguity problem. However, the overconfident choice of the output-label assignment by PIT results in a sub-optimal trained model. In this work, we propose a probabilistic optimization framework to address the inefficiency of PIT in finding the best output-label assignment. Our proposed method entitled trainable Soft-minimum PIT is then employed on the same Long-Short Term Memory (LSTM) architecture used in Permutation Invariant Training (PIT) speech separation method. The results of our experiments show that the proposed method outperforms conventional PIT speech separation significantly (p-value $ < 0.01$) by +1dB in Signal to Distortion Ratio (SDR) and +1.5dB in Signal to Interference Ratio (SIR).

Impact of Dynamic LED Non-Linearity on DCO-OFDM Optical Wireless Communication

DC-Offset Orthogonal Frequency Division Multiplexing (DCO-OFDM) is a popular modulation method for Optical Wireless Communication (OWC) as it can assign specific power levels and constellations to different subcarrier frequencies across the low-pass response of the LED. However, LEDs exhibit dynamic non-linearities to which OFDM is very sensitive. As LED signals are subject to non-linear memory effects in photon generation, different OFDM subcarriers experience different amounts of distortion. This letter translates the LED behavior into a communication channel model for OFDM modulation. It gives an experimentally verified expression for the Signal-to-Distortion Ratio (SDR) as it varies per subcarrier in OFDM systems.

Hardware design for blind source separation using fast time-frequency mask technique

In this paper, we propose a fast time-frequency mask technique that relies on the sparseness of source signals for blind source separation (BSS) to separate a mixture of two input sounds in a single signal automatically. Due to the sparseness of source signals, the signal can be distinguished when it is transformed into the time-frequency domain. Most previous methods did not mention the effect of different angles on accuracy. To overcome such problems, we first define two features which are normalized level-ratio and phase-difference. Next, we use our method to decrease the variance of Direction of Arrival (DOA). This can reduce the variance of features so that it can reduce the iterations of k-means. Finally, a mask is generated according to the clustered features. Our method does not require any prior information or parameter estimation. The motivation of the proposed design is to incorporate the BSS system with some smart voice appliances. In the application scenario, all the non-human voices may appear and regard as interference. We use Signal to Distortion Ratio (SDR) and Signal to Interference Ratio (SIR) to make some comparison. Based on the proposed system, then we present a hardware design. We use the TSMC 90-nm CMOS process. As a cost-effective result, it consumes about 120 K gates and executes with a frequency of 10 MHz. The power consumption is only 2.92 mW with low power design considerations.

An Efficient Multistage Approach for Blind Source Separation of Noisy Convolutive Speech Mixture

This paper proposes a novel efficient multistage algorithm to extract source speech signals from a noisy convolutive mixture. The proposed approach comprises two stages named Blind Source Separation (BSS) and de-noising. A hybrid source prior model separates the source signals from the noisy reverberant mixture in the BSS stage. Moreover, we model the low- and high-energy components by generalized multivariate Gaussian and super-Gaussian models, respectively. We use Minimum Mean Square Error (MMSE) to reduce noise in the noisy convolutive mixture signal in the de-noising stage. Furthermore, the two proposed models investigate the performance gain. In the first model, the speech signal is separated from the observed noisy convolutive mixture in the BSS stage, followed by suppression of noise in the estimated source signals in the de-noising module. In the second approach, the noise is reduced using the MMSE filtering technique in the received noisy convolutive mixture at the de-noising stage, followed by separation of source signals from the de-noised reverberant mixture at the BSS stage. We evaluate the performance of the proposed scheme in terms of signal-to-distortion ratio (SDR) with respect to other well-known multistage BSS methods. The results show the superior performance of the proposed algorithm over the other state-of-the-art methods.