Single-channel Speech Signal Research Articles

Comparative Study on Noisy Speech Preprocessing Algorithms

Speech Enhancement and noise reduction have wide applications in speech processing. They are often employed as pre-processing stage in various applications. The work to be presented in this paper is denoising a single-channel speech signal at the presence of a highly non-stationary background noise in order to improve the perceptible quality and intelligibility of the speech. Real world noise is mostly highly non-stationary and does not affect the speech signal uniformly over the spectrum. This paper investigates various Discrete Fourier Transform-based algorithms as single-channel pre-processing techniques consisting of: Spectral Subtraction using over-subtraction and spectral floor, Multi-Band Spectral Subtraction (MBSS), Wiener Filter, MMSE of Short-Time Spectral Amplitude (MMSE-STSA) estimator with, and without using SPU modifier, MMSE Log-Spectral Amplitude Estimator with, and without using SPU modifier, Optimally-Modified Log-Spectral Amplitude estimator (OM-LSA). The processed speeches from these algorithms are compared at the same set of conditions using visual examinations of signals in the time domain and the spectrograms, and also the objective and subjective tests for quality and perceptual evaluation. All the implemented algorithms provide considerable, different degrees of flexibility and control on noise elimination levels that reduces artifacts in the enhanced speech, resulting in the improved quality, and intelligibility.

Layered Convolutive Nonnegative Matrix Factorization for Speech Separation

Abstract Nonnegative matrix factorization (NMF) has attracted significant attention for its good performance in single-channel speech separation. The improved algorithms of NMF have become research hotspots. Layered NMF (LNMF), an improved algorithm, can express the source signal more accurately for its multilayer structure. However, LNMF sometimes performs poorly because it ignores the short-term correlation of speech signals. Based on LNMF and the advantages of Convolutive NMF (CNMF), we proposed a Layered Convolutive NMF(LCNMF) algorithm for single-channel speech separation. The LCNMF corporates the multilayer structure into the NMF and expands the convolution of the top-level NMF model. During the training, NMF is used to learn the non-top-level basis matrices, and CNMF is used to learn the top-level basis matrix, then combined with each single-layer of basis matrix. During the prediction, CNMF is used to separate mixed signals. The results on the dataset MIK-1K showed that LCNMF outperformed NMF and LNMF for separating the mixture of single-channel speech signals. LCNMF improved by 0.019, 1.049dB, 1.305dB, and 0.851dB on average compared with NMF, and improved by 0.007, 0.172dB, 0.090dB, and 0.366dB on average compared with LNMF in sort-term objective intelligibility (STOI), Source to Distortion Ratio (SDR), Source to Interference Ratio (SIR) and Source to Artifacts Ratio (SAR)

Adaptive Speech Spectrogram Approximation for Enhancement of Speech Signal

The process of speech enhancement tends to decrease the noise with keeping undistorted speech signal amplitude. There are several benefits of speech processing systems which comes with the some challenges. In this paper, we proposed adaptive speech spectrogram approximation (ASSA) technique that used to tackle the de-noising and dereverberation in a single channel speech signal. The model is processed using sparse representation prototype in order to perform the de-noising process, where it remove the noise that present in speech signal more thoroughly.Where matrix factorization and SIFT is used to model the speech signal spectrogram, a time-varying filter is used to minimalize the noise more effectively.The noise adaptive model is implemented via iterative updating parameters in order to approximate the lower reverberant speech signal in a SIFT domain.Afterwards, the proposed ASSA technique compute the variation in estimated speech signal in order to decrease the noise components and to predict the final speech magnitude. In order to evaluate the performance of proposed system it is compared with state-of-art techniques using some performance metrics.

Joint Estimation of Reverberation Time and Early-To-Late Reverberation Ratio From Single-Channel Speech Signals

The reverberation time (RT) and the early-to-late reverberation ratio (ELR) are two key parameters commonly used to characterize acoustic room environments. In contrast to conventional blind estimation methods that process the two parameters separately, we propose a model for joint estimation to predict the RT and the ELR simultaneously from single-channel speech signals from either full-band or sub-band frequency data, which is referred to as joint room parameter estimator (jROPE). An artificial neural network is employed to learn the mapping from acoustic observations to the RT and the ELR classes. Auditory-inspired acoustic features obtained by temporal modulation filtering of the speech time-frequency representations are used as input for the neural network. Based on an in-depth analysis of the dependency between the RT and the ELR, a two-dimensional (RT, ELR) distribution with constrained boundaries is derived, which is then exploited to evaluate four different configurations for jROPE. Experimental results show that-in comparison to the single-task ROPE system which individually estimates the RT or the ELR-jROPE provides improved results for both tasks in various reverberant and (diffuse) noisy environments. Among the four proposed joint types, the one incorporating multi-task learning with shared input and hidden layers yields the best estimation accuracies on average. When encountering extreme reverberant conditions with RTs and ELRs lying beyond the derived (RT, ELR) distribution, the type considering RT and ELR as a joint parameter performs robustly, in particular. From state-of-the-art algorithms that were tested in the acoustic characterization of environments challenge, jROPE achieves comparable results among the best for all individual tasks (RT and ELR estimation from full-band and sub-band signals).

Blind reverberation energy estimation using exponential averaging with attack and release time constants for hearing aids

Dereverberation signal processing is necessary for hearing-aid systems because reverberation degrades speech intelligibility in reverberant environments. In general, hearing-aid systems require low computational cost and real-time signal processing. The spectral subtraction (SS) method is a simple and frequently used technique which is used not only for noise-reduction but for dereverberation as well. To perform dereverberation methods based on SS for hearing aids, the reverberation energy must be estimated or measured. In this paper, the SS-based blind estimation of reverberation energy in a single-channel speech signal is proposed by using exponential averaging with attack and release time constants. The estimation error, which is the difference between the true and estimated reverberation energy, was used for evaluation. The estimation error of the proposed method was compared with the results of the method proposed by Lebart et al., which is a well-known non-blind SS-based dereverberation method. According to the results, the reverberation energy was more accurately estimated when the reverberation time was longer than 0.6 s, and the estimation error of the proposed method was approximately half of that of the well-known non-blind method by Lebart.

Experimental Study on Extreme Learning Machine Applications for Speech Enhancement

In wireless telephony and audio data mining applications, it is desirable that noise suppression can be made robust against changing noise conditions and operates in real time (or faster). The learning effectiveness and speed of artificial neural networks are therefore critical factors in applications for speech enhancement tasks. To address these issues, we present an extreme learning machine (ELM) framework, aimed at the effective and fast removal of background noise from a single-channel speech signal, based on a set of randomly chosen hidden units and analytically determined output weights. Because feature learning with shallow ELM may not be effective for natural signals, such as speech, even with a large number of hidden nodes, hierarchical ELM (H-ELM) architectures are deployed by leveraging sparse autoencoders. In this manner, we not only keep all the advantages of deep models in approximating complicated functions and maintaining strong regression capabilities, but we also overcome the cumbersome and time-consuming features of both greedy layer-wise pre-training and back-propagation (BP)-based fine tuning schemes, which are typically adopted for training deep neural architectures. The proposed ELM framework was evaluated on the Aurora-4 speech database. The Aurora-4 task provides relatively limited training data, and test speech data corrupted with both additive noise and convolutive distortions for matched and mismatched channels and signal-to-noise ratio (SNR) conditions. In addition, the task includes a subset of testing data involving noise types and SNR levels that are not seen in the training data. The experimental results indicate that when the amount of training data is limited, both ELMand H-ELM-based speech enhancement techniques consistently outperform the conventional BP-based shallow and deep learning algorithms, in terms of standardized objective evaluations, under various testing conditions.

An Efficient Speaker Recognition System for Separating the Single Channel Speech Using Frequency Modulation

Separating multiple speech signals from single channel sound signal made a huge challenge between researchers. In our paper, we intended a Modulation Frequency technique for Single channel Speech separation. Speech is perceived as having a pitch at the fundamental frequency of vibration of the vocal chords. The combined speech is first decomposed into a two-dimensional time-frequency representation using Short-time Fourier analysis (STFT) with frequency bins. It will pass through a narrow band pass Filter. A cross channel correlation did not use its fundamental frequency directly and while it uses tools to estimate pitch. A frequency bins contain fundamental frequency will correlate with the frequency bins with harmonics and the preset value to be initiated. A standard speaker identification system based on Gaussian Mixture models is of frame basis to identify the interfering speaker. A two parallel signal separation, one based on correlations of modulation frequency and the other based on fundamental frequency alone approaches is obtained. Our proposed method is implemented in MATLAB and verified using various speech signals. The experimental results show that the proposed technique is more efficient to separate the multiple voices from Single Channel Speech Signal.

Kalman‐filters in subbands for noise reduction with enhanced pitch‐adaptive speech model estimation

AbstractIn this paper, we present an approach for a Kalman filter which enhances single channel speech signals disturbed by car noise. For the Kalman filters, the speech and noise signals are modelled by AR processes and the corresponding models have to be estimated, based on the noisy speech. To avoid high model orders, the filtering is performed in subbands. We especially focus on the estimation of the speech model: First, standard model estimation methods are investigated and the method giving the best stable model is then utilized as a preliminary solution. To enhance the models without generating musical tones, a new method is proposed that utilizes estimates of the speaker's pitch frequency. The algorithm we present has less than 39 ms delay and thus fulfills the ETSI requirements for mobile telephony [1].