Search

Saliency-driven image and video coding for humans has gained importance in the recent past. In this paper, we propose such a saliency-driven coding framework for the video coding for machines task using the latest video coding standard Versatile Video Coding (VVC). To determine the salient regions before encoding, we employ the real-time-capable object detection network You Only Look Once~(YOLO) in combination with a novel decision criterion. To measure the coding quality for a machine, the state-of-the-art object segmentation network Mask R-CNN was applied to the decoded frame. From extensive simulations we find that, compared to the reference VVC with a constant quality, up to 29 % of bitrate can be saved with the same detection accuracy at the decoder side by applying the proposed saliency-driven framework. Besides, we compare YOLO against other, more traditional saliency detection methods., Comment: 5 pages, 3 figures, 2 tables; Originally submitted at IEEE ICASSP 2021

We propose a novel reinforcement learning approach to learning lattice codes for MIMO channels. We use the block error rate as a loss function to be minimized and compare the learnt lattices with those obtained from algebraic design methods for different SNR ranges. Our results indicate that our learnt lattices achieve close to optimal performance in some cases.

Lossless compression of datasets is a problem of significant theoretical and practical interest. It appears naturally in the task of storing, sending, or archiving large collections of information for scientific research. We can greatly improve encoding bitrate if we allow the compression of the original dataset to decompress to a permutation of the data. We prove the equivalence of dataset compression to compressing a permutation-invariant structure of the data and implement such a scheme via predictive coding. We benchmark our compression procedure against state-of-the-art compression utilities on the popular machine-learning datasets MNIST and CIFAR-10 and outperform for multiple parameter sets.

Scene background initialization is an important step as one low-layer method for high-layer applications in computer vision. However, this process is always affected by practical challenges such as illumination changes, back-ground motion, camera jitter, intermittent movement and bad weather outdoors, etc. In this work, we develop a novel method called co-occurrence pixel-block (CPB) model via spatial-temporal correlation for robust back-ground initialization. This work first introduces the CPB method for foreground extraction. And then, background information in spatial-temporal features are utilized to recover an adaptive background for the current frame. Experimental results obtained from the dataset of the challenging benchmark (SBMnet) validate it’s performance under various challenges.

In the automatic speech recognition (ASR) system, how to solve the problem of code-switch speech recognition has been a concern. Code-switch speech recognition is challenging due to data scarcity as well as diverse syntactic structures across languages. In this paper, we focus on the code-switch speech recognition in mainland China, which is obviously different from the Hong Kong and Southeast Asia area in linguistic characteristics. We propose a novel approach that only uses monolingual data for code-switch second-pass speech recognition which is also named language model rescoring. The approach converts the code-switch sentence to a monolingual sentence by a word mapping and language model determination step, therefore the issue of data scarcity is unnecessary to be considered. The word pairs during the word mapping step are generated by a fine-designed generation process that incorporates machine translation, word alignment, etc. We show that the proposed approach achieves an over 7.23% relative WER reduction from the naive monolingual language model (MLM) rescoring in our test set.

There is a fundamental trade-off between the channel representation resolution of codebooks and the overheads of feedback communications in the fifth generation new radio (5G NR) frequency division duplex (FDD) massive multiple-input and multiple-output (MIMO) systems. In particular, two types of codebooks (namely Type I and Type II codebooks) are introduced with different resolution and overhead requirements. Although the Type I codebook based scheme requires lower feedback overhead, its channel state information (CSI) reconstruction and beamforming performance are not as good as those from the Type II codebook based scheme. However, since the Type I codebook based scheme has been widely used in 4G systems for many years, replacing it by the Type II codebook based scheme overnight is too costly to be an option. Therefore, in this paper, using Type I codebook, we leverage advances in cutting plane method to optimize the CSI reconstruction at the base station (BS), in order to close the gap between these two codebook based beamforming schemes. Numerical results based on channel samples from QUAsi Deterministic RadIo channel GenerAtor (QuaDRiGa) are presented to show the excellent performance of the proposed algorithm in terms of beamforming vector acquisition.

In this work, we present a simplification and a corresponding hardware architecture for hard-decision recursive projectionaggregation (RPA) decoding of Reed-Muller (RM) codes. In particular, we transform the recursive structure of RPA decoding into a simpler and iterative structure with minimal error-correction degradation. Our simulation results for RM(7; 3) show that the proposed simplification has a small error-correcting performance degradation (0:005 in terms of channel crossover probability) while reducing the average number of computations by up to 40%. In addition, we describe the first fully parallel hardware architecture for simplified RPA decoding. We present FPGA implementation results for an RM(6; 3) code on a Xilinx Virtex-7 FPGA showing that our proposed architecture achieves a throughput of 171 Mbps at a frequency of 80 MHz.

Consistent in-focus input imagery is an essential precondition for machine vision systems to perceive the dynamic environment. A de-focus blur severely degrades the performance of vision systems. To tackle this problem, we propose a deep-learning-based framework estimating the kernel scale and orientation of the defocus blur to ad-just lens focus rapidly. Our pipeline utilizes 3D ConvNet for a variable number of input hypotheses to select the optimal slice from the input stack. We use random shuffle and Gumbel-softmax to improve network performance. We also propose to generate synthetic defocused images with various asymmetric coded apertures to facilitate training. Experiments are conducted to demonstrate the effectiveness of our framework.

The recent emergence of machine-learning based generative models for speech suggests a significant reduction in bit rate for speech codecs is possible. However, the performance of generative models deteriorates significantly with the distortions present in real-world input signals. We argue that this deterioration is due to the sensitivity of the maximum likelihood criterion to outliers and the ineffectiveness of modeling a sum of independent signals with a single autoregressive model. We introduce predictive-variance regularization to reduce the sensitivity to outliers, resulting in a significant increase in performance. We show that noise reduction to remove unwanted signals can significantly increase performance. We provide extensive subjective performance evaluations that show that our system based on generative modeling provides state-of-the-art coding performance at 3 kb/s for real-world speech signals at reasonable computational complexity.

In AOMedia Video 1 (AV1), directional intra prediction modes are applied to model local texture patterns that present certain directionality. Each intra prediction direction is represented with a nominal mode index and a delta angle. The delta angle is entropy coded using shared context between luma and chroma, and the context is derived using the associated nominal mode. In this paper, two methods are proposed to further reduce the signaling cost of delta angles: cross-component delta angle coding, and context-adaptive delta angle coding, whereby the cross-component and spatial correlation of the delta angles are explored, respectively. The proposed methods were implemented on top of a recent version of libaom. Experimental results show that the proposed cross-component delta angle coding achieved average 0.4% BD-rate reduction with 4% encoding time saving over all intra configurations. By combining both methods, an average 1.2% BD-rate reduction is achieved.

Hyperspectral unmixing is an important tool to learn the material constitution and distribution of a scene. Model-based unmixing methods depend on well-designed iterative optimization algorithms, which is usually time consuming. Learning-based methods perform unmixing in a data-driven manner but heavily rely on the quality and quantity of the training samples due to the lack of physical interpretability. In this paper, we combine the advantages of both model-based and learning-based methods and propose a nonnegative matrix factorization (NMF) inspired sparse autoencoder (NMF-SAE) for hyperspectral unmixing. NMF-SAE consists of an encoder and a decoder, both of which are constructed by unrolling the iterative optimization rules of L 1 sparsity-constrained NMF for the linear spectral mixture model. All parameters in our method are obtained by end-to-end training in a data-driven manner. Our network is not only physically interpretable and flexible but also has higher learning capacity with fewer parameters. Experimental results on both synthetic and real-world data demonstrate that our method is capable of producing desirable unmixing results when compared against several alternative approaches.

We live in a world where 60% of the population can speak two or more languages fluently. Members of these communities constantly switch between languages when having a conversation. As automatic speech recognition (ASR) systems are being deployed to the real-world, there is a need for practical systems that can handle multiple languages both within an utterance or across utterances. In this paper, we present an end-to-end ASR system using a transformer-transducer model architecture for code-switched speech recognition. We propose three modifications over the vanilla model in order to handle various aspects of code-switching. First, we introduce two auxiliary loss functions to handle the low-resource scenario of code-switching. Second, we propose a novel mask-based training strategy with language ID information to improve the label encoder training towards intra-sentential code-switching. Finally, we propose a multi-label/multi-audio encoder structure to leverage the vast monolingual speech corpora towards code-switching. We demonstrate the efficacy of our proposed approaches on the SEAME dataset, a public Mandarin-English code-switching corpus, achieving a mixed error rate of 18.5% and 26.3% on test_man and test_sge sets respectively., Comment: Accepted at ICASSP 2021

We consider a semantic communication system for speech signals, named DeepSC-S. Motivated by the breakthroughs in deep learning (DL), we make an effort to recover the transmitted speech signals in the semantic communication systems, which minimizes the error at the semantic level rather than the bit level or symbol level as in the traditional communication systems. Particularly, based on an attention mechanism employing squeeze-and-excitation (SE) networks, we design the transceiver as an end-to-end (E2E) system, which learns and extracts the essential speech information. Furthermore, in order to facilitate the proposed DeepSC-S to work well on dynamic practical communication scenarios, we find a model yielding good performance when coping with various channel environments without retraining process. The simulation results demonstrate that our proposed DeepSC-S is more robust to channel variations and outperforms the traditional communication systems, especially in the low signal-to-noise (SNR) regime., 6 pages. arXiv admin note: text overlap with arXiv:2107.11190

The system combining i-vector and probabilistic linear discriminant analysis (PLDA) has been applied with great success in the speaker recognition task. The i-vector space gives a low-dimensional representation of a speech segment and training data of a PLDA model, which offers greater robustness under different conditions. In this paper, we propose a new framework based on i-vector/PLDA and Maximum Entropy (ME) to improve the performance of speaker identification system in the presence of speech coding distortion. The results are reported on TIMIT database and speech coding obtained by passing the speech test from TIMIT database through the AMR encoder/decoder. Our results show that the proposed methode achieves improved performance when compared with the i-vector/PLDA and MEGMM.

Adaptive multi-rate (AMR), a popular audio compression standard, is widely used in mobile communication and mobile Internet applications and has become a novel carrier for hiding information. To improve the statistical security, this paper presents a steganographic scheme in the AMR fixed codebook (FCB) domain based on the pulse distribution model (PDM-AFS), which is obtained from the distribution characteristics of the FCB value in the cover audio. The pulse positions in stego audio are controlled by message encoding and random masking to make the statistical distribution of the FCB parameters close to that of the cover audio. The experimental results show that the statistical security of the proposed scheme is better than that of the existing schemes. Furthermore, the hiding capacity is maintained compared with the existing schemes. The average hiding capacity can reach 2.06 kbps at an audio compression rate of 12.2 kbps, and the auditory concealment is good. To the best of our knowledge, this is the first secure AMR FCB steganographic scheme that improves the statistical security based on the distribution model of the cover audio. This scheme can be extended to other audio compression codecs under the principle of algebraic code excited linear prediction (ACELP), such as G.723.1 and G.729.

We propose a speech coding scheme based on the simple transform coding and forward adaptive quantization for discrete input signal processing in this paper. The quasi-logarithmic quantizer is applied to discretization of continuous input signal, i.e. for preparing discrete input. The application of forward adaptation based on the input signal variance provides more efficient bandwidth usage, whereas utilization of transform coding provides sub-sequences with more predictable signal characteristics that ensure higher quality of signal reconstruction at the receiving end. In order to provide additional compression, transform coding precedes adaptive quantization. As an objective measure of system performance we use signal-to-quantization-noise ratio. Sysem performance is discussed for two typical cases. In the first case, we consider that the information about continuous signal variance is available whereas the second case considers system performance estimation when we know only the information about discretized signal variance which means that there is a loss of input signal information. The main goal of such performance estimation comparison of the proposed speech signal coding model is to explore what is the objectivity of performance if we do not have information about a continuous source, which is a common phenomenon in digital systems.

A new end-to-end communication system is proposed to increase transmission speed, robustness, and security in order to meet the requirements of mobile systems that know an exponentially increasing data amount over time. The design relies on the use of compressed sensing-source coding instead of the supported speech coding standards in actual mobile communication systems. The proposed compressed sensing-source coding method allows reducing the speech coding complexity by using simple quantisation and binary encoding, saving communication system resources, and encrypting communications without additional costs. The performance of the resulting communication system is evaluated for speech communication via 10 dB Rayleigh environment in terms of perceptual evaluation of speech quality (PESQ) scores and coherence speech intelligibility index (CSII) when convolutional coding, orthogonal frequency division multiplexing, and diversity schemes are used. Results report that for a bit rate of 12.8 kbit/s the proposed scheme achieves fair speech intelligibility justified by a CSII value of 0.5, and offers good output speech quality measure, providing a PESQ of 3.33 for the same bit rate.

This paper investigates an approach for speaker identification in a remote access system based on coded speech signals. The aim of using the coding process is to decrease the amount of transmitted data via the channel. In this proposed system, the speech signal is coded by two different coding techniques. It can be coded either by linear predictive coding or compressive sensing. The coded speech signal is transmitted into the receiver via the wireless communication channel. At the receiver, the received signal is decoded, and then speaker identification system is applied on the decoded signal. During the transmission process, the channel errors affect on the transmitted signal, so they should be taken into account. The speaker identification process is used to achieve the security needed for the remote access system. In speaker identification system, the feature vectors are captured from different discrete transforms such as discrete wavelet transform, discrete cosine transform, and discrete sine transform, besides the time domain. The recognition rate for all transforms is computed to evaluate the effect of coded signals on the performance of the speaker identification system. The results proved that the discrete cosine transform and discrete wavelet transform are the best. In addition to the proposed system gives close recognition results to those obtained from real speech signals revealing a simple degradation effect due to the speech coding.

Global system for mobile communication (GSM) is widely used digital mobile service around the world. Although GSM was designed as a secure wireless system, it is now vulnerable to different targeted attacks. There is a need to address security domains especially the confidentiality of communication. This paper presents a novel framework for end to end secure voice communication over the GSM networks using encryption algorithm AES-256. A special Modem and speech coding technique are designed to enable the transmission of encrypted speech using GSM voice channel. To the best of our knowledge, this is first solution that uses single codebook for transmission of secure voice. An efficient low bit-rate (1.9 kbps) speech coder is also designed for use with the proposed modulation scheme for optimal results. Different speech characteristics such as pitch, energy and line spectral frequencies are extracted and preserved before compression and encryption of speech. Previously, the best achieved data rate was 1.6 kbps with three codebooks, whilst the proposed approach achieves 2 kbps with 0% bit error rate. The empirical results show that the methodology can be used for real time applications to transmit encrypted voice using GSM network.

Speech recognition was measured in 24 normal-hearing subjects for unprocessed speech and for speech processed by a cochlear implant Advanced Combination Encoder (ACE) coding strategy in quiet and at various signal-to noise ratios (SNRs). All signals were low- or high-pass filtered to avoid ceiling effects. Surprisingly, speech recognition performance plateaus at approximately 22 dB SNR for both speech types, implying that ACE processing has no effect on the upper limit of the effective SNR range. Speech recognition improved significantly above 15 dB SNR, suggesting that the upper limit used in the Speech Intelligibility Index should be reconsidered.

In this study, the authors present a new steganographic technique called random least significant bits of pitch and Fourier magnitude steganography (RLPFS). It is based on hiding a secret speech coded by mixed excitation linear prediction (MELP) speech coder in speech bitstream (cover signal), which is also encoded by MELP coder. First, the RLPFS leaks the hidden speech in the following modes: pitch-based steganography, Fourier magnitude-based steganography or both. These modes are selected randomly. Second, during transmission, the stego speech, the mode number, and the number of embedded bits would be transmitted either through a covert channel created in the transmission protocol or through the cover speech. In this work, the authors have dealt with the challenge of embedding a secret speech into a cover speech coded by a very low bit rate speech coder while maintaining a reasonable level of speech quality. They have shown that RLPFS was able to create hidden channels with maximum steganographic bandwidths up to 266.64 bit/s at the cost of a steganographic noise between 0.031 and 0.62 mean opinion score. Also, this study takes into account the security of the parameters, the synchronisation of the receiver to deal with a packet loss during transmission and the resistance of the proposed method against steganalysis.

Vector quantization (VQ) is a very effective way to save bandwidth and storage for speech coding and image coding. Traditional vector quantization methods can be divided into mainly seven types, tree-structured VQ, direct sum VQ, Cartesian product VQ, lattice VQ, classified VQ, feedback VQ, and fuzzy VQ, according to their codebook generation procedures. Over the past decade, quantization-based approximate nearest neighbor (ANN) search has been developing very fast and many methods have emerged for searching images with binary codes in the memory for large-scale datasets. Their most impressive characteristics are the use of multiple codebooks. This leads to the appearance of two kinds of codebook: the linear combination codebook and the joint codebook. This may be a trend for the future. However, these methods are just finding a balance among speed, accuracy, and memory consumption for ANN search, and sometimes one of these three suffers. So, finding a vector quantization method that can strike a balance between speed and accuracy and consume moderately sized memory, is still a problem requiring study.

Conventional narrowband (NB) telephony suffers from limited acoustic bandwidth at the receiver side, leading to degraded speech quality and intelligibility. In this paper, artificial speech bandwidth extension (ABE) of NB speech toward missing frequencies below about 300 Hz (low-frequency band, LB) is proposed to enhance the speech quality. The LB-ABE in this paper is employed together with a preexisting ABE toward high-frequency components to obtain spectrally balanced speech signals. In an instrumental quality assessment, the spectral distance in the LB was improved by more than 5 dB compared to NB speech. In a subjective listening test, the gap of speech quality between wideband and NB speech was significantly reduced when employing the proposed ABE toward low frequencies. The LB extension was found to further improve the preexisting ABE toward higher frequencies by a significant 0.26 CMOS points.

Delta Modulation (DM) is a simple waveform coding algorithm used mostly when timely data delivery is more important than the transmitted data quality. While the implementation of DM is fairly simple and inexpensive, it suffers from several limitations, such as slope overload and granular noise, which can be overcome using Adaptive Delta Modulation (ADM). This paper presents novel 2-digit ADM with six-level quantization using variable-length coding, for encoding the time-varying signals modelled by Laplacian distribution. Two variants of quantizer are employed, distortion-constrained quantizer which is optimally designed for minimal mean-squared error (MSE), and rate-constrained quantizer, which is suboptimal in the minimal MSE sense, but enables minimal loss in SQNR for the target bit rate. Experimental results using real speech signal are provided, indicating that the proposed configuration outperforms the baseline ADM algorithms, including Constant Factor Delta Modulation (CFDM), Continuously Va...

This paper proposes a convolutional neural network (CNN)-based encoder model to compress and code speech signal directly from raw input speech. Although the model can synthesize wideband speech by implicit bandwidth extension, narrowband is preferred for IP telephony and telecommunications purposes. The model takes time domain speech samples as inputs and encodes them using a cascade of convolutional filters in multiple layers, where pooling is applied after some layers to downsample the encoded speech by half. The final bottleneck layer of the CNN encoder provides an abstract and compact representation of the speech signal. In this paper, it is demonstrated that this compact representation is sufficient to reconstruct the original speech signal in high quality using the CNN decoder. This paper also discusses the theoretical background of why and how CNN may be used for end-to-end speech compression and coding. The complexity, delay, memory requirements, and bit rate versus quality are discussed in the experimental results. Web of Science 7 75089 75081

Voice-over IP (VoIP) technology is a kind of digital transmission technology based on IP network. It is one of the important methods to use voice service in VoIP as steganographic carrier to ensure secure transmission. However, the traditional steganographic code has some problems, such as low embedding efficiency and weak concealment, which cannot meet the requirements of VoIP streaming media information hiding for the security of secret information. Therefore, a steganographic algorithm combining F5 and simplified wet paper code (SWPC) algorithm is proposed. The main idea is to embed secret messages in each row of the carrier matrix using the F5 algorithm, and then the SWPC algorithm is used to embed the columns according to the wet and dry characteristics of the wet paper code without affecting the results before row embedding. We use the VoIP streams encoded by the ITU-T G.729a codec as a carrier to verify the proposed scheme. The experimental results demonstrate that the proposed scheme can achieve relatively better IP speech data steganographic transparency and that it can outperform F5-WPC and SWPC approaches.

Purpose: Many objective measures are developed to assess speech in specific contexts such as speech coding, transmission, etc. Speech enhancement is one of emergent speech technology applications. Yet, there is no objective standard to assess denoised speech. Indeed, the objective criteria are developed mainly to assess speech for specific context i.e., speech coding, speech transmission, etc. They are not developed specially to assess denoised speech. Some attempts are developed in the literature based on a linear combination of existing measures. Method: Even such approaches permit a promising performances, they reach their limits. To overcome these limits, we propose to address the problem from the point of view of classification instead of the point of view of regression. The assessed speech utterance is divided into frames. Each frame is evaluated using one of conventional criteria. Hence, a feature vector is constructed. The objective denoising speech assessment becomes a classical classification problem. Results: Unlike traditional criteria, the proposed method can give a significant objective score directly interpreted as an estimation of real mean opinion score. Conclusion: It is shown in this study that is possible to predict the subjective mean opinion score with acceptable and fairly sensitivity, specificity, precision and accuracy.

Compared with more and more steganography techniques motivated by abundant compressed speech, steganalysis is still a challenging task. Many existing studies are based on a single dimensional feature model and it is difficult to have a wide range of applicability. In this paper, a hybrid Markov model is proposed, which is based on the correlation of fixed codebook parameters in speech codec between pulses in a given track. And then, two detecting methods based on entropy are given. One is designed as a single-pulse position based entropy detection method (SPBE). The other is pulse-position pairs based entropy detection method (PPBE). Simultaneously, to solve the problem of inaccurate calculation of the entropy rate of finite length samples, corrected conditional entropy (CCE) is used as an estimate of the Markov chain entropy rate. Experiments show that CCE and entropy are highly complementary, and both can be employed as classification features to achieve better steganalysis results. Finally, the performance of the proposed detection methods is evaluated and compared with the existing detection methods. Results prove that the two methods proposed in this paper are suitable for online and real-time steganographic detection, especially for small-size samples.

Musical training is associated with a myriad of neuroplastic changes in the brain, including more robust and efficient neural processing of clean and degraded speech signals at brainstem and cortical levels. These assumptions stem largely from cross-sectional studies between musicians and nonmusicians which cannot address whether training itself is sufficient to induce physiological changes or whether preexisting superiority in auditory function before training predisposes individuals to pursue musical interests and appear to have similar neuroplastic benefits as musicians. Here, we recorded neuroelectric brain activity to clear and noise-degraded speech sounds in individuals without formal music training but who differed in their receptive musical perceptual abilities as assessed objectively via the Profile of Music Perception Skills. We found that listeners with naturally more adept listening skills (“musical sleepers”) had enhanced frequency-following responses to speech that were also more resilient to the detrimental effects of noise, consistent with the increased fidelity of speech encoding and speech-in-noise benefits observed previously in highly trained musicians. Further comparisons between these musical sleepers and actual trained musicians suggested that experience provides an additional boost to the neural encoding and perception of speech. Collectively, our findings suggest that the auditory neuroplasticity of music engagement likely involves a layering of both preexisting (nature) and experience-driven (nurture) factors in complex sound processing. In the absence of formal training, individuals with intrinsically proficient auditory systems can exhibit musician-like auditory function that can be further shaped in an experience-dependent manner.

The Advanced Combination Encoder (ACE) signal processing strategy is used in the majority of cochlear implant (CI) sound processors manufactured by Cochlear Corporation. This “n-of-m” strategy selects “n” out of “m” available frequency channels with the highest spectral energy in each stimulation cycle. It is hypothesized that at low signal-to-noise ratio (SNR) conditions, noise–dominant frequency channels are susceptible for selection, neglecting channels containing target speech cues. In order to improve speech segregation in noise, explicit encoding of formant frequency locations within the standard channel selection framework of ACE is suggested. Two strategies using the direct formant estimation algorithms are developed within this study, FACE (formant-ACE) and VFACE (voiced-activated-formant-ACE). Speech intelligibility from eight CI users is compared across 11 acoustic conditions, including mixtures of noise and reverberation at multiple SNRs. Significant intelligibility gains were observed with VFACE over ACE in 5 dB babble noise; however, results with FACE/VFACE in all other conditions were comparable to standard ACE. An increased selection of channels associated with the second formant frequency is observed for FACE and VFACE. Both proposed methods may serve as potential supplementary channel selection techniques for the ACE sound processing strategy for cochlear implants.

This article investigates switched quantizers for speech signal depicted with Gaussian probability density function (PDF). Gaussian PDF is better for smaller frame lengths that are represented here. Companding technique results in constant Signal to Quantization Noise Ratio (SQNR). In this paper two approaches are present: quasi-logarithmic (QL) and piecewise uniform (PU) compandor. Simpler compandor directly affects the complexity of hardware realization and expense of given solution, but, on the other hand, also brings to weaker performances. Therefore, a smart choice has to be made. Usage of switched technique leads to better performances. This way, the quality of quantization is improved by dividing the dynamic range of variances into multiple subranges. For each subrange a separate quantizer is designed, with its support region’s amplitude. The optimal amplitude is numerically determined, whereby a single criterion is obtaining the maximal SQNR. Bit rates of these quantizers don’t depend on signal variance, as the fixed length codes are used. The performances of proposed quantizers are demonstrated on real speech signals from the reliable database. Comparison of obtained results with other recent solutions is done in order to show the efficiency of this model. DOI: http://dx.doi.org/10.5755/j01.eie.24.6.22295

In wireless communication, speech is an essential service for which low rate speech coding is a key technique. Reducing the transmission bandwith and achieving higher speech quality are primary concerns in developing new speech coding algorithms. This report develops a joint channel and source controlled adaptive variable bit-rate algorithm to achieve a high speech quality and maintain an efficient spectrum.

As a challenging pattern recognition task, speech emotion recognition has attracted more and more attention in recent years and is widely used in medical, Affective Computing, and other fields. In this paper, we proposed a parallel network of ResNet-CNN-Transformer Encoder. The Res-Net is used to alleviate the problems caused by the deepening of the network. The CNN calculates the fewer parameters to increase the fitting expression ability of the network. Due to the traditional recurrent neural network, with a long-term dependence on the feature extraction of speech and text sequences and sequence attributes not capturing long-distance features, the multi attention mechanism of the transformer coding layer is used to parallelize the sequence, improve the processing speed and extract the emotional semantic information in the sequence. Experiments are carried out on the RAVDESS dataset. Our results demonstrate the effectiveness of the proposed method and make a significant improvement compared with the previous results.

Interpolation is a common operation in many applications, like communications, measurers, speech coding, radars, navigation systems, etc. As the information technologies advances, new fields of applications for the interpolation as a required operation to perform appears. The problem of an accurate choice of the interpolation method for specific application is relevant today. In the paper, to provide an effective reconstruction of periodic discrete-time signals in practical applications, five key requirements for the interpolation method that must be satisfied are presented. It is concerned that there is no interpolation method nowadays to be able to fully satisfy the mentioned requirements. Analysis of the sampling theorem generalizations for periodic signals is performed. A novel interpolation algorithm for the arbitrary finite duration periodic signals reconstruction is considered. The algorithm is based on using of the finite-length basis interpolation functions ones derived from the sinc-function. The advantages and disadvantaged of the new algorithm are listed. Comparison of the proposed algorithm and the existed signal reconstruction methods is performed.

A brief review of low-rate coding methods with the inclusion of neural networks in the codec structure is carried out. The issue of using radial neural networks for solving problems of low-speed coding and reducing the energy consumption of information transmission systems without communication quality degradation. The structure of a codec using radial neural networks in Matlab code is presented. Simulation of three-channel transmission system was performed for two signal type: human voice and music. Codec structure for each signal is the same, but neural network for music processing contains 69 neurons in hidden layer, while for voice 27 neurons is enough. The simulation shows that the application of a radial neural network as part of a predictive codec can reduce the dynamic range of signals by up to 70% while maintaining the communication quality for each of the used types of signal. At the same time, the simulation shows that codec with neural networks causes additional latency in signal processing.

The performance of speech coding, speech recognition, and speech enhancement largely depends upon the accuracy of the linear prediction coefficient (LPC) of clean speech and noise in practice. Formulation of speech and noise LPC estimation as a supervised learning problem has shown considerable promise. In its simplest form, a supervised technique, typically a deep neural network (DNN) is trained to learn a mapping from noisy speech features to clean speech and noise LPCs. Training targets for DNN to clean speech and noise LPC estimation fall into four categories: line spectrum frequency (LSF), LPC power spectrum (LPC-PS), power spectrum (PS), and magnitude spectrum (MS). The choice of appropriate training target as well as the DNN method can have a significant impact on LPC estimation in practice. Motivated by this, we perform a comprehensive study on the training targets using two state-of-the-art DNN methods--- residual network and temporal convolutional network (ResNet-TCN) and multi-head attention network (MHANet). This study aims to determine which training target as well as DNN method produces more accurate LPCs in practice. We train the ResNet-TCN and MHANet for each training target with a large data set. Experiments on the NOIZEUS corpus demonstrate that the LPC-PS training target with MHANet produces a lower spectral distortion (SD) level in the estimated speech LPCs in real-life noise conditions. We also construct the AKF with the estimated speech and noise LPC parameters from each training target using ResNet-TCN and MHANet. Subjective AB listening tests and seven different objective quality and intelligibility evaluation measures (CSIG, CBAK, COVL, PESQ, STOI, SegSNR, and SI-SDR) on the NOIZEUS corpus demonstrate that the AKF constructed with MHANet-LPC-PS driven speech and noise LPC parameters produced enhanced speech with higher quality and intelligibility than competing methods.

In order to further improve the performance of speech bandwidth extension, this paper proposes an end-to-end codec network model for speech bandwidth extension. It is modeled on the time-domain waveform and consists of encoder network, long short-term memory (LSTM) and decoder network. The encoder network is responsible for feature extraction and data dimensionality reduction of the input data, the LSTM is responsible for extracting the context-dependent information of the speech signal, and the decoder network is responsible for wideband speech reconstruction of the potential features output by the LSTM. In addition, this paper also proposes a time-frequency perception loss function to guide model training to generate more accurate time-domain waveforms and more realistic frequency-domain spectrum. Experimental results show that the reconstructed wideband speech generated by the model achieves better results in both subjective and objective evaluation.