Search

This paper proposes a static modeling and power flow calculation of the More Electric Aircraft (MEA) power system to study the stability and the fault analysis. Combined with the common land power system, the structure and characteristics of MEA power system are firstly illustrated and the static models are built. Then, on the basis of the alternate iteration method, the method to calculate and analysis the AC/DC hybrid MEA power system is described. Finally, the power flow calculation results and adjust strategy under different working period are showed with an example of Boeing 787. The simulation results showed that the programmed algorithm can accord with the characteristics of MEA power system and the power flow distribution and adjustment strategy are simple and clear.

Space vector PWM inverter has been emerged as a promising technique to design a powerful supply for AC motor control applications in modern motor drive systems. In this paper, we implement a hardware model of three-phase voltage source inverter based on space vector PWM algorithm using 16-bit Digital Signal Controller dsPIC30F4011. The experimental results are carefully investigated to demonstrate that our proposed approach gains a high performance.

The learners' interaction and contents in online discussion forum could also be as important elements to evaluate the learning. The paper proposes a framework of concept map derived from forum (CMF) to analyze online discussion forum in Moodle. The framework analyzes the interaction among peers and posts, which facilitates effective knowledge organization by constructing relationships between forum contents and existing knowledge. The framework extracts the learning difficulties and visualized with word clouds in online forum. The research contributes to bridging the gap between teaching materials and students' questions from online forum. The CMF framework is helpful to design the architecture of automatically answer system and real-time feedback within adaptive learning in the future.

This paper proposes a design method of the feed- forward controller to improve on the non-interference control performance of a single-phase AC/DC PWM converter in parallel for high speed trains. The feed-forward controller compensates interference phenomenon generated by switching converter in parallel. As the secondary leakage inductance in transformer is changed by operating conditions of the trans- former, the gain value of feed-forward controller is calculated by estimating the changed leakage inductance. The estimation procedure of the boost inductance is only based on the variation of input current and the input AC voltage measurement. The estimated boost inductance is optimized by the least square method. Then the controller with gain value can exactly compensate the interference phenomenon. The validity of proposed technique is verified through the MATLAB Simulink simulation results.

Electromagnetic interference can cause disruptions to the functional signal on a wired communication channel. Detecting whether the transmitted data has been affected by electromagnetic interference is an important challenge that needs to be addressed. This paper describes an advanced design of an electromagnetic interference detector to detect electromagnetic interference in a wired communication channel, and compares it with the previously proposed adder and subtractor based A&S electromagnetic interference detector. The performance of both EMI detectors is evaluated in light of recently proposed electromagnetic interference condition monitoring definitions. These definitions categorize the received signal based on the impact of electromagnetic disturbance on the transmitted data and its detection by the EMI detector, which helps to determine if the overall detection is contributing towards enhancing safety and/or availability. An in-depth numerical analysis of both electromagnetic interference detectors is performed using a mathematical model. The mathematical model considers different scenarios for the electromagnetic disturbance that is present at the location of the electromagnetic interference detector. These scenarios include varying amplitude, frequency, and phase as well as varying phase differences of the voltages induced by the electromagnetic disturbance in each of the data transmission lines. In addition, the performance of both electromagnetic interference detectors is analyzed in semi-reverberant environments using a Monte-Carlo simulation framework. This research concludes that the Advanced electromagnetic interference detector addresses the inadequacies of the A&S electromagnetic interference detector and can be effectively used for enhancing the safety of wired communication channels. ispartof: Ieee Access status: Published online

Remote attestation (RA) is a primitive that allows the authentication of software components on untrusted systems by relying on a root of trust. Network protocols can use the primitive to establish trust in remote software components they communicate with. As such, RA can be regarded as a first-class security primitive like (a)symmetric encryption, message authentication, etc. However, current formal models of RA do not allow analysing protocols that use the primitive without tying them to specific platforms, low-level languages, memory protection models, or implementation details. In this paper, we propose and demonstrate a new model, called pi_RA, that supports RA at a high level of abstraction by treating it as a cryptographic primitive in a variant of the applied pi-calculus. To demonstrate the use of pi_RA, we use it to formalise and analyse the security of MAGE, an SGX-based framework that allows mutual attestation of multiple enclaves. The protocol is formalised in the form of a compiler that implements actor-based communication primitives in a source language (pi_Actor) in terms of remote attestation primitives in pi_RA. Our security analysis uncovers a caveat in the security of MAGE that was left unmentioned in the original paper. ispartof: pages:284-298 ispartof: 2023 IEEE 36th Computer Security Foundations Symposium (CSF) pages:284-298 ispartof: 2023 IEEE 36th Computer Security Foundations Symposium (CSF) location:Dubrovnik, Croatia date:9 Jul - 13 Jul 2023 status: published

This paper presents a novel, highly compact direct- injection modular universal power flow and quality control topology exclusively using lower power components. In addition to conventional high-voltage applications, it is particularly attractive for the distribution grid down to low voltage as it can exploit the latest developments in low-voltage high-current semiconductors. The distribution grid is under immense pressure due to rapid growth of high-power loads and sources, such as vehicle chargers and photovoltaic installations, which can locally overload lines and generate both over- and under-voltage conditions in the same grid section so that conventional voltage adjustment through transformer tap changers fails. In contrast to conventional universal power flow and/or quality controllers (UPFC, UPQC) or soft-open point (SOP) circuits, the concept both eliminates any injection transformer, which is large, costly, and limited in dynamics and the need to convert the full grid power. The presented direct-injection circuit exploits the recent developments in high-current low-voltage transistors. The centre piece is a low-voltage module connected in series with each phase, which is floating with the grid voltage so that it only needs to deal with a small voltage difference between its outputs and can maintain that condition also under fault currents of several kiloampères. Due to the low-impedance of the distribution grid, only a few volts are sufficient to control several hundred ampères in active and reactive current in a loop. In contrast to commercially available transformer-injection-based SOP solutions with often considerable additional loss, the injection modules have almost negligible additional impedance (< 4 mΩ).

In recent years, data mining had become very popular to provide a facility in various fields. Large amount of data elements are stored in the database. Various data mining algorithms are used to classify the data elements and perform some operations to find the best solution on the basis of some relational parameters. One of the automatic architecture designs in this research is to find the solution for examination in terms of level wise question paper. The entire examination process is a vital component for direct assessment of an individual learning. So, preparing a complete test paper and the setup is relatively necessary. Currently, the customary technique of making question paper has been handbook. In current scenario the question paper generation is a manual approach leading to unproductive at times owing to bias, repetition and security concerns. The current paper presents an automatic procedure of question paper group which can be modified, streamlined, synchronized and secured. Each task done by this system is automatic, such that storing space, bias and security is not an apprehension any longer. Earlier, the question paper was generated by concerned subject teacher manually and was very time consuming, man power was required and sometimes the question paper lacked accuracy. Outcome Based Education (OBE) designates what students will know and be intelligent to do, as they advance in a program. The information related to student learning was collected for question paper generation and assessment using OBE. This information can be used to predict students ability, advancements in education system, betterment in teaching method, future interest of student etc. Various data mining classifiers like oneR, ZeroR, J48, Naive Bayes, IBk are used for prediction of the course outcome. The comparative study contains some parameters like time, detection accuracy, classification error etc. to find the performance of the proposed architecture as compared to other techniques to verify the enhancements.

This paper studies the integration of two successful hardware-supported security mechanisms: capabilities and enclaved execution. Capabilities are a powerful and flexible security mechanism for implementing fine-grained memory access control and compartmentalizing untrusted or buggy software components. Capabilities have a long history but have gained significant momentum recently, as evidenced by ARM’s experimental Morello processor that supports the Capability Hardware Enhanced RISC Instructions (CHERI). Enclaved execution is a popular mechanism for dynamically creating Trusted Execution Environments (TEEs), called enclaves. Enclaves are isolated execution contexts that protect the integrity and confidentiality of software in the enclave (even against compromised system software) and that support attestation. Integrating capabilities and enclaved execution in a single processor is challenging because they overlap partially in their security objectives, and a clean integration should unify the way in which these overlapping objectives are achieved. In addition, it is not obvious how attestation should interact with capabilities. In this paper, we propose CHERI-TrEE: a novel design for a processor that cleanly integrates support for both capabilities and enclaved execution. CHERI-TrEE targets low-end embedded systems without virtual memory. We show that CHERI-TrEE is greater than the sum of its parts by showing how it naturally supports useful features that have traditionally been hard to support in enclaved execution, like dynamically growing and shrinking enclaves, non-contiguous and nested enclaves, sharing of memory between enclaves etc. We implement our proposal both in hardware on a RISC-V processor, as well as in a small software hypervisor on top of ARM Morello, and evaluate impact on performance and hardware resources. ispartof: pages:1143-1159 ispartof: Proceedings of the 8th IEEE European Symposium on Security and Privacy pages:1143-1159 ispartof: 8th IEEE European Symposium on Security and Privacy location:Delft date:3 Jul - 7 Jun 2023 status: published

This paper presents a novel shape descriptor to effectively and efficiently characterize the local image statistics. The proposed descriptor, termed contour covariance (CC), characterizes covariance features driven by a moving point on the shape contour at multiple scales. To calculate the covariance matrices, three basic features including texture, intensity and distance map, are extracted from the object image. Based on coefficients of the obtained covariance matrices, the proposed CC descriptor is compact yet informative, as well as invariant to rotation, translation and scale. The experimental results on two databases demonstrate the superiority and efficiency of the proposed method among the state-of-the-art methods for shape classification.

A novel combination forecasting model is presented in this paper, which combines single ones based on machine learning. The model has been applied to the prediction of five cities' election in Taiwan with combining the exposure rate and the approval rate, which obtains good results. The exposure rate is the frequency of a candidate's appearances in the news and approval rate is the proportion of the positive information of a candidate. And the polarity of a review is predicted by sentiment classification based on machine learning techniques. A novel method of feature extraction is used in sentiment classification, which makes the classifier effectively assign the review a type of polarity. Meanwhile, this paper proposes a method of feature clustering and extending based on the synonym dictionary, which obviously reduces the dimension of feature vector and improve the F-score of sentiment classification.

This Innovative Practice Work in Progress presents progress in developing exercises for high school students incorporating level-appropriate mathematics into robotics activities. We assume mathematical foundations ranging from algebra to precalculus, whereas most prior work on integrating mathematics into robotics uses only very elementary mathematical reasoning or, at the other extreme, is comprised of technical papers or books using calculus and other advanced mathematics. The exercises suggested are relevant to any differential-drive robot, which is an appropriate model for many different varieties of educational robots. They guide students towards comparing a variety of natural navigational strategies making use of typical movement primitives. The exercises align with Common Core State Standards for Mathematics.

This Innovative Practice Work in Progress paper presents a experience of internship assessment in two bachelor’s degrees in Business administration, Industrial Engineering and Computer Engineering in Spain. In both studies internship are compulsory and although those studies have in essence many differences, in fact they have many common elements because both are orientated to professionalizing students. These characteristics encourage the authors, who are involved in internship as academic tutors, to propose an objective assessment model that can be used by different higher education degrees as evaluation model for internship. Therefore, this work defines a comprehensive evaluation process based on competencies for external academic internships at higher education. Integral in the sense of the participation of the academic and company tutors; and working in a double field: the academic and the professional. The evaluation model is based on the design of an evaluation rubric; from which emerges a set of assessment and qualification documents to be used as a reference/guide to shows the level of competencies mastery by each student and how they will be measured. The final of the assessment process is an individualized Final report that summarizes student’s learning process and his/her final quantitative rating; This report also includes the comments of the tutors about the skills worked, the progress in every competency and the different levels acquired by the student.

For the realization of future wireless networks, gigabit wireless technology which can achieve higher-than-1Gbps data transmission with extremely low transmit power is indispensable. We have been studying the distributed antenna network (DAN) and the frequency-domain wireless signal processing. In DAN, many antennas or clusters of antennas are spatially distributed over a service area and they are connected by means of optical fiber links with DAN signal processing center (SPC). A number of distributed antennas cooperatively serve mobile users using spatial multiplexing, diversity, array or relaying technique. In this paper, the recent research advances of gigabit DAN and its related wireless techniques are introduced.

A promising approach to improve the efficiency in manufacturing processes is to build digital twins of the main assets from the available data to estimate and predict important metrics such as the energy consumption of each machine. Such predictions can then be exploited in process optimization. However, generating the digital twins from data is challenging in Industry 4.0: the high dimensionality of the data and the complexity of the manufacturing process makes it difficult and time-consuming for a human to model all important assets in the line. To address this problem, this paper investigates the feasibility of a framework for automatic generation of digital twins from historical data, which incorporates automatic selection of variables and data-driven system identification. For that, different versions of an algorithm for variable selection were exhaustively combined with several data-driven system identification methods. The framework was applied in the generation of digital twins for an electrical motor in a state-of-the-art production line of wooden fibreboards. In the estimation and prediction of the motor power, several of the automatically-generated digital twins presented acceptable accuracy, performing similarly to a simple manually designed physics-based model. The results suggest that the automatic framework is feasible, deserving therefore further effort to enhance it and provide a more extensive validation. ispartof: International Conference on Control, Automation and Diagnosis ispartof: 2023 International Conference on Control, Automation and Diagnosis (ICCAD) location:Rome, Italy date:10 May - 12 May 2023 status: Published online

This paper presents a traffic signal scheduling strategy for a pedestrian-vehicle mixed-flow network with consideration of the pedestrian two-way crossing phase (TWC) and the exclusive pedestrian phase (EPP) in the urban traffic system. According to the pedestrian crossing characteristics, two volume dynamics are firstly proposed for the TWC phase setting and EPP phase setting, respectively. Secondly, a mathematical model combining both the vehicle traffic network and the pedestrian flows at each intersection is developed, with the aim to minimize the delay cost from two different traffic participants (pedestrians and vehicles) and to provide suggestions on selecting EPP or TWC phase pattern. Numerical simulation results are provided to illustrate the effectiveness of our real-time traffic light scheduling strategy and the potential impact of two phase patterns when different traffic volumes are involved in the system.

Analog memory offers enormous potential to speed up computation in deep learning. We study the use of phase-change memory (PCM) as the resistive element in a crossbar array that allows the multiply-accumulate operation in deep neural networks to be performed in-memory. With this promise comes several challenges, including this paper’s main focus: the impact of conductance drift on deep neural network accuracy. Here we offer an overview of our recent work, including explanations of popular neural network architectures, along with a technique to compensate for drift ("slope correction") to allow in-memory computing with PCM during inference to reach software-equivalent deep learning baselines for a broad variety of important neural network workloads.

In recent years, multi-level steganography has become a new research direction in the network steganography. In order to ensure transmitting the key and the ciphertext covertly in multi-level steganography, a solution of key transmission is proposed in this paper. The proposed solution consists of 2 methods. The upper-level method is used to transmit the ciphertext. The lower-level is used to deliver the key covertly. The experimental results showed that this solution can realize synchronous transmission of the ciphertext and the key, which has a high steganographic bandwidth and a low steganographic cost, improving the security of network steganography.

Cortico-muscular communication patterns reveal important informa- tion about motor control. However, inferring significant causal re- lationships between motor cortex electroencephalogram (EEG) and surface electromyogram (sEMG) of concurrently active muscles is challenging since relevant processes involved in muscle control are relatively weak compared to additive noise and background activi- ties. In this paper, a framework for identification of cortico-muscular linear time invariant communication is proposed that simultaneously estimates model order and its parameters by enforcing sparsity and stationarity conditions in a convex optimization program. The exper- imental results demonstrate that our proposed algorithm outperforms existing techniques for autoregressive model estimation, in terms of computational speed and model identification for causality estima- tion.

Object trackers deployed on low-power devices need to be light-weight, however, most of the current state-of-the-art (SOTA) methods rely on using compute-heavy backbones built using CNNs or transformers. Large sizes of such models do not allow their deployment in low-power conditions and designing compressed variants of large tracking models is of great importance. This paper demonstrates how highly compressed light-weight object trackers can be designed using neural architectural pruning of large CNN and transformer based trackers. Further, a comparative study on architectural choices best suited to design light-weight trackers is provided. A comparison between SOTA trackers using CNNs, transformers as well as the combination of the two is presented to study their stability at various compression ratios. Finally results for extreme pruning scenarios going as low as 1% in some cases are shown to study the limits of network pruning in object tracking. This work provides deeper insights into designing highly efficient trackers from existing SOTA methods., Comment: Accepted at IEEE ICASSP 2023

Bayesian Federated Learning (FL) offers a principled framework to account for the uncertainty caused by limitations in the data available at the nodes implementing collaborative training. In Bayesian FL, nodes exchange information about local posterior distributions over the model parameters space. This paper focuses on Bayesian FL implemented in a device-to-device (D2D) network via Decentralized Stochastic Gradient Langevin Dynamics (DSGLD), a recently introduced gradient-based Markov Chain Monte Carlo (MCMC) method. Based on the observation that DSGLD applies random Gaussian perturbations of model parameters, we propose to leverage channel noise on the D2D links as a mechanism for MCMC sampling. The proposed approach is compared against a conventional implementation of frequentist FL based on compression and digital transmission, highlighting advantages and limitations., submitted

In this paper, we extend previous self-supervised approaches for language identification by experimenting with Conformer based architecture in a multilingual pre-training paradigm. We find that pre-trained speech models optimally encode language discriminatory information in lower layers. Further, we demonstrate that the embeddings obtained from these layers are significantly robust to classify unseen languages and different acoustic environments without additional training. After fine-tuning a pre-trained Conformer model on the VoxLingua107 dataset, we achieve results similar to current state-of-the-art systems for language identification. More, our model accomplishes this with 5x less parameters. We open-source the model through the NVIDIA NeMo toolkit., Comment: Submitted to ICASSP 2023

This paper introduces topological Slepians, i.e., a novel class of signals defined over topological spaces (e.g., simplicial complexes) that are maximally concentrated on the topological domain (e.g., over a set of nodes, edges, triangles, etc.) and perfectly localized on the dual domain (e.g., a set of frequencies). These signals are obtained as the principal eigenvectors of a matrix built from proper localization operators acting over topology and frequency domains. Then, we suggest a principled procedure to build dictionaries of topological Slepians, which theoretically provide non-degenerate frames. Finally, we evaluate the effectiveness of the proposed topological Slepian dictionary in two applications, i.e., sparse signal representation and denoising of edge flows., Comment: 6 pages, 3 figures

Language Models today provide a high accuracy across a large number of downstream tasks. However, they remain susceptible to adversarial attacks, particularly against those where the adversarial examples maintain considerable similarity to the original text. Given the multilingual nature of text, the effectiveness of adversarial examples across translations and how machine translations can improve the robustness of adversarial examples remain largely unexplored. In this paper, we present a comprehensive study on the robustness of current text adversarial attacks to round-trip translation. We demonstrate that 6 state-of-the-art text-based adversarial attacks do not maintain their efficacy after round-trip translation. Furthermore, we introduce an intervention-based solution to this problem, by integrating Machine Translation into the process of adversarial example generation and demonstrating increased robustness to round-trip translation. Our results indicate that finding adversarial examples robust to translation can help identify the insufficiency of language models that is common across languages, and motivate further research into multilingual adversarial attacks., Comment: Published at International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2023

Enhancing speech quality is an indispensable yet difficult task as it is often complicated by a range of degradation factors. In addition to additive noise, reverberation, clipping, and speech attenuation can all adversely affect speech quality. Speech restoration aims to recover speech components from these distortions. This paper focuses on exploring the impact of self-supervised speech representation learning on the speech restoration task. Specifically, we employ speech representation in various speech restoration networks and evaluate their performance under complicated distortion scenarios. Our experiments demonstrate that the contextual information provided by the self-supervised speech representation can enhance speech restoration performance in various distortion scenarios, while also increasing robustness against the duration of speech attenuation and mismatched test conditions., To be presented at ICASSP 2023

Multi-task learning aims to acquire a set of functions, either regressors or classifiers, that perform well for diverse tasks. At its core, the idea behind multi-task learning is to exploit the intrinsic similarity across data sources to aid in the learning process for each individual domain. In this paper we draw intuition from the two extreme learning scenarios -- a single function for all tasks, and a task-specific function that ignores the other tasks dependencies -- to propose a bias-variance trade-off. To control the relationship between the variance (given by the number of i.i.d. samples), and the bias (coming from data from other task), we introduce a constrained learning formulation that enforces domain specific solutions to be close to a central function. This problem is solved in the dual domain, for which we propose a stochastic primal-dual algorithm. Experimental results for a multi-domain classification problem with real data show that the proposed procedure outperforms both the task specific, as well as the single classifiers.

This paper describes an end-to-end (E2E) neural architecture for the audio rendering of small portions of display content on low resource personal computing devices. It is intended to address the problem of accessibility for vision-impaired or vision-distracted users at the hardware level. Neural image-to-text (ITT) and text-to-speech (TTS) approaches are reviewed and a new technique is introduced to efficiently integrate them in a way that is both efficient and back-propagate-able, leading to a non-autoregressive E2E image-to-speech (ITS) neural network that is efficient and trainable. Experimental results are presented showing that, compared with the non-E2E approach, the proposed E2E system is 29% faster and uses 19% fewer parameters with a 2% reduction in phone accuracy. A future direction to address accuracy is presented., 5 pages, 3 figures, 2023 IEEE Conference on Acoustics, Speech, and Signal Processing

Filter pruning has attracted increasing attention in recent years for its capacity in compressing and accelerating convolutional neural networks. Various data-independent criteria, including norm-based and relationship-based ones, were proposed to prune the most unimportant filters. However, these state-of-the-art criteria fail to fully consider the dissimilarity of filters, and thus might lead to performance degradation. In this paper, we first analyze the limitation of relationship-based criteria with examples, and then introduce a new data-independent criterion, Weighted Hybrid Criterion (WHC), to tackle the problems of both norm-based and relationship-based criteria. By taking the magnitude of each filter and the linear dependence between filters into consideration, WHC can robustly recognize the most redundant filters, which can be safely pruned without introducing severe performance degradation to networks. Extensive pruning experiments in a simple one-shot manner demonstrate the effectiveness of the proposed WHC. In particular, WHC can prune ResNet-50 on ImageNet with more than 42% of floating point operations reduced without any performance loss in top-5 accuracy., Accepted by ICASSP 2023

Shadow removal improves the visual quality and legibility of digital copies of documents. However, document shadow removal remains an unresolved subject. Traditional techniques rely on heuristics that vary from situation to situation. Given the quality and quantity of current public datasets, the majority of neural network models are ill-equipped for this task. In this paper, we propose a Transformer-based model for document shadow removal that utilizes shadow context encoding and decoding in both shadow and shadow-free regions. Additionally, shadow detection and pixel-level enhancement are included in the whole coarse-to-fine process. On the basis of comprehensive benchmark evaluations, it is competitive with state-of-the-art methods.

Noisy Student Training (NST) has recently demonstrated extremely strong performance in Automatic Speech Recognition(ASR). In this paper, we propose a data selection strategy named LM Filter to improve the performance of NST on non-target domain data in ASR tasks. Hypotheses with and without a Language Model are generated and the CER differences between them are utilized as a filter threshold. Results reveal that significant improvements of 10.4% compared with no data filtering baselines. We can achieve 3.31% CER in AISHELL-1 test set, which is best result from our knowledge without any other supervised data. We also perform evaluations on the supervised 1000 hour AISHELL-2 dataset and competitive results of 4.73% CER can be achieved., This paper is accepted by the ICASSP 2023 conference

In this paper we propose a novel generative approach, DiffRoll, to tackle automatic music transcription (AMT). Instead of treating AMT as a discriminative task in which the model is trained to convert spectrograms into piano rolls, we think of it as a conditional generative task where we train our model to generate realistic looking piano rolls from pure Gaussian noise conditioned on spectrograms. This new AMT formulation enables DiffRoll to transcribe, generate and even inpaint music. Due to the classifier-free nature, DiffRoll is also able to be trained on unpaired datasets where only piano rolls are available. Our experiments show that DiffRoll outperforms its discriminative counterpart by 19 percentage points (ppt.) and our ablation studies also indicate that it outperforms similar existing methods by 4.8 ppt. Source code and demonstration are available https://sony.github.io/DiffRoll/.

Transformer-based models attain excellent results and generalize well when trained on sufficient amounts of data. However, constrained by the limited data available in the audio domain, most transformer-based models for audio tasks are finetuned from pre-trained models in other domains (e.g. image), which has a notable gap with the audio domain. Other methods explore the self-supervised learning approaches directly in the audio domain but currently do not perform well in the downstream tasks. In this paper, we present a novel self-supervised learning method for transformer-based audio models, called masked spectrogram prediction (MaskSpec), to learn powerful audio representations from unlabeled audio data (AudioSet used in this paper). Our method masks random patches of the input spectrogram and reconstructs the masked regions with an encoder-decoder architecture. Without using extra model weights or supervision, experimental results on multiple downstream datasets demonstrate MaskSpec achieves a significant performance gain against the supervised methods and outperforms the previous pre-trained models. In particular, our best model reaches the performance of 0.471 (mAP) on AudioSet, 0.854 (mAP) on OpenMIC2018, 0.982 (accuracy) on ESC-50, 0.976 (accuracy) on SCV2, and 0.823 (accuracy) on DCASE2019 Task1A respectively., Submit to INTERSPEECH 2022

The success of the multilingual automatic speech recognition systems empowered many voice-driven applications. However, measuring the performance of such systems remains a major challenge, due to its dependency on manually transcribed speech data in both mono- and multilingual scenarios. In this paper, we propose a novel multilingual framework -- eWER3 -- jointly trained on acoustic and lexical representation to estimate word error rate. We demonstrate the effectiveness of eWER3 to (i) predict WER without using any internal states from the ASR and (ii) use the multilingual shared latent space to push the performance of the close-related languages. We show our proposed multilingual model outperforms the previous monolingual word error rate estimation method (eWER2) by an absolute 9\% increase in Pearson correlation coefficient (PCC), with better overall estimation between the predicted and reference WER., Comment: Accepted in ICASSP, Multilingual WER estimation, End-to-End systems, multilingual model, automatic word error rate estimation

Acoustic signal processing in the spherical harmonics domain (SHD) is an active research area that exploits the signals acquired by higher order microphone arrays. A very important task is that concerning the localization of active sound sources. In this paper, we propose a simple yet effective method to localize prominent acoustic sources in adverse acoustic scenarios. By using a proper normalization and arrangement of the estimated spherical harmonic coefficients, we exploit low-rank approximations to estimate the far field modal directional pattern of the dominant source at each time-frame. The experiments confirm the validity of the proposed approach, with superior performance compared to other recent SHD-based approaches., Comment: To appear in ICASSP 2023

In recent years, the task of Automatic Music Transcription (AMT), whereby various attributes of music notes are estimated from audio, has received increasing attention. At the same time, the related task of Multi-Pitch Estimation (MPE) remains a challenging but necessary component of almost all AMT approaches, even if only implicitly. In the context of AMT, pitch information is typically quantized to the nominal pitches of the Western music scale. Even in more general contexts, MPE systems typically produce pitch predictions with some degree of quantization. In certain applications of AMT, such as Guitar Tablature Transcription (GTT), it is more meaningful to estimate continuous-valued pitch contours. Guitar tablature has the capacity to represent various playing techniques, some of which involve pitch modulation. Contemporary approaches to AMT do not adequately address pitch modulation, and offer only less quantization at the expense of more model complexity. In this paper, we present a GTT formulation that estimates continuous-valued pitch contours, grouping them according to their string and fret of origin. We demonstrate that for this task, the proposed method significantly improves the resolution of MPE and simultaneously yields tablature estimation results competitive with baseline models., Comment: Accepted to ICASSP 2023

Federated learning (FL) is a popular technique for training a global model on data distributed across client devices. Like other distributed training techniques, FL is susceptible to straggler (slower or failed) clients. Recent work has proposed to address this through device-to-device (D2D) offloading, which introduces privacy concerns. In this paper, we propose a novel straggler-optimal approach for coded matrix computations which can significantly reduce the communication delay and privacy issues introduced from D2D data transmissions in FL. Moreover, our proposed approach leads to a considerable improvement of the local computation speed when the generated data matrix is sparse. Numerical evaluations confirm the superiority of our proposed method over baseline approaches., Comment: arXiv admin note: text overlap with arXiv:2301.12685

End-to-end (E2E) automatic speech recognition (ASR) implicitly learns the token sequence distribution of paired audio-transcript training data. However, it still suffers from domain shifts from training to testing, and domain adaptation is still challenging. To alleviate this problem, this paper designs a replaceable internal language model (RILM) method, which makes it feasible to directly replace the internal language model (LM) of E2E ASR models with a target-domain LM in the decoding stage when a domain shift is encountered. Furthermore, this paper proposes a residual softmax (R-softmax) that is designed for CTC-based E2E ASR models to adapt to the target domain without re-training during inference. For E2E ASR models trained on the LibriSpeech corpus, experiments showed that the proposed methods gave a 2.6% absolute WER reduction on the Switchboard data and a 1.0% WER reduction on the AESRC2020 corpus while maintaining intra-domain ASR results., Comment: Accepted by ICASSP2023

The emotion detection technology to enhance human decision-making is an important research issue for real-world applications, but real-life emotion datasets are relatively rare and small. The experiments conducted in this paper use the CEMO, which was collected in a French emergency call center. Two pre-trained models based on speech and text were fine-tuned for speech emotion recognition. Using pre-trained Transformer encoders mitigates our data's limited and sparse nature. This paper explores the different fusion strategies of these modality-specific models. In particular, fusions with and without cross-attention mechanisms were tested to gather the most relevant information from both the speech and text encoders. We show that multimodal fusion brings an absolute gain of 4-9% with respect to either single modality and that the Symmetric multi-headed cross-attention mechanism performed better than late classical fusion approaches. Our experiments also suggest that for the real-life CEMO corpus, the audio component encodes more emotive information than the textual one., Comment: 5 pages, 2 figures, 4 tables

International audience; The speaker diarization task answers the question "who is speaking at a given time?". It represents valuable information for scene analysis in a domain such as robotics. In this paper, we introduce a temporal audiovisual fusion model for multiusers speaker diarization, with low computing requirement, a good robustness and an absence of training phase. The proposed method identifies the dominant speakers and tracks them over time by measuring the spatial coincidence between sound locations and visual presence. The model is generative, parameters are estimated online, and does not require training. Its effectiveness was assessed using two datasets, a public one and one collected in-house with the Pepper humanoid robot.

Transformer-based methods have shown great potential in long-term time series forecasting. However, most of these methods adopt the standard point-wise self-attention mechanism, which not only becomes intractable for long-term forecasting since its complexity increases quadratically with the length of time series, but also cannot explicitly capture the predictive dependencies from contexts since the corresponding key and value are transformed from the same point. This paper proposes a predictive Transformer-based model called {\em Preformer}. Preformer introduces a novel efficient {\em Multi-Scale Segment-Correlation} mechanism that divides time series into segments and utilizes segment-wise correlation-based attention for encoding time series. A multi-scale structure is developed to aggregate dependencies at different temporal scales and facilitate the selection of segment length. Preformer further designs a predictive paradigm for decoding, where the key and value come from two successive segments rather than the same segment. In this way, if a key segment has a high correlation score with the query segment, its successive segment contributes more to the prediction of the query segment. Extensive experiments demonstrate that our Preformer outperforms other Transformer-based methods.

The sixth-generation networks envision the terahertz (THz) band as one of the key enabling technologies because of its ultrawide bandwidth. To combat the severe attenuation, the THz wireless systems employ large arrays, wherein the near-field beam-split (NB) severely degrades the accuracy of channel acquisition. Contrary to prior works that examine only either narrowband beamforming or far-field models, we estimate the wideband THz channel via an NB-aware orthogonal matching pursuit (NBA-OMP) approach. We design an NBA dictionary of near-field steering vectors by exploiting the corresponding angular and range deviation. Our OMP algorithm accounts for this deviation thereby ipso facto mitigating the effect of NB. Numerical experiments demonstrate the effectiveness of the proposed channel estimation technique for wideband THz systems., Accepted Paper in 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

The use of mutual information as a tool in private data sharing has remained an open challenge due to the difficulty of its estimation in practice. In this paper, we propose InfoShape, a task-based encoder that aims to remove unnecessary sensitive information from training data while maintaining enough relevant information for a particular ML training task. We achieve this goal by utilizing mutual information estimators that are based on neural networks, in order to measure two performance metrics, privacy and utility. Using these together in a Lagrangian optimization, we train a separate neural network as a lossy encoder. We empirically show that InfoShape is capable of shaping the encoded samples to be informative for a specific downstream task while eliminating unnecessary sensitive information. Moreover, we demonstrate that the classification accuracy of downstream models has a meaningful connection with our utility and privacy measures., 5 pages, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Human-robot interaction relies on a noise-robust audio processing module capable of estimating target speech from audio recordings impacted by environmental noise, as well as self-induced noise, so-called ego-noise. While external ambient noise sources vary from environment to environment, ego-noise is mainly caused by the internal motors and joints of a robot. Ego-noise and environmental noise reduction are often decoupled, i.e., ego-noise reduction is performed without considering environmental noise. Recently, a variational autoencoder (VAE)-based speech model has been combined with a fully adaptive non-negative matrix factorization (NMF) noise model to recover clean speech under different environmental noise disturbances. However, its enhancement performance is limited in adverse acoustic scenarios involving, e.g. ego-noise. In this paper, we propose a multichannel partially adaptive scheme to jointly model ego-noise and environmental noise utilizing the VAE-NMF framework, where we take advantage of spatially and spectrally structured characteristics of ego-noise by pre-training the ego-noise model, while retaining the ability to adapt to unknown environmental noise. Experimental results show that our proposed approach outperforms the methods based on a completely fixed scheme and a fully adaptive scheme when ego-noise and environmental noise are present simultaneously., Comment: Accepted to the 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2023)

Recently, a relative transfer function (RTF)-vector-based method has been proposed to estimate the direction of arrival (DOA) of a target speaker for a binaural hearing aid setup, assuming the availability of external microphones. This method exploits the external microphones to estimate the RTF vector corresponding to the binaural hearing aid and constructs a one-dimensional spatial spectrum by comparing the estimated RTF vector against a database of anechoic prototype RTF vectors for several directions. In this paper we assume the availability of a calibrated array of external microphones, which is characterized by a second database of anechoic prototype RTF vectors. We propose a method, where the external microphones are not only exploited to estimate the RTF vector corresponding to the binaural hearing aid but also assist in estimating the DOA of the target speaker. Based on the estimated RTF vector for all microphones and both prototype databases, a two-dimensional spatial spectrum is constructed from which the DOA is estimated. Experimental results for a reverberant environment with diffuse-like noise show that assisted DOA estimation outperforms DOA estimation where the prototype database characterizing the array of external microphones is not used., Submitted to ICASSP 2023

We present design considerations for an automated meeting scheduling agent that processes meeting requests on behalf of its associated user. In our formulation of the meeting scheduling problem, distributed meeting scheduling agents, one per user, intelligently exchange information with each other to schedule meetings without compromising user-specified constraints. In this paper, we first enumerate various strategies we have investigated to focus distributed negotiation between scheduling agents. Next, we demonstrate the necessity for such a scheduler to be adaptive in its choice of options for the various strategy dimensions, so that it can perform effectively over time. In order to build an adaptive scheduler that can effectively choose from available strategy options, we develop quantitative performance estimates of these options using detailed probabilistic analysis. Results from these analyses are used to provide guidelines to choose the most appropriate strategy combination given current environmental conditions and local problem-solving states. >

Meetings are an essential form of communication for all types of organizations, and remote collaboration systems have been much more widely used since the COVID-19 pandemic. One major issue with remote meetings is that it is challenging for remote participants to interrupt and speak. We have recently developed the first speech interruption analysis model, which detects failed speech interruptions, shows very promising performance, and is being deployed in the cloud. To deliver this feature in a more cost-efficient and environment-friendly way, we reduced the model complexity and size to ship the WavLM_SI model in client devices. In this paper, we first describe how we successfully improved the True Positive Rate (TPR) at a 1% False Positive Rate (FPR) from 50.9% to 68.3% for the failed speech interruption detection model by training on a larger dataset and fine-tuning. We then shrank the model size from 222.7 MB to 9.3 MB with an acceptable loss in accuracy and reduced the complexity from 31.2 GMACS (Giga Multiply-Accumulate Operations per Second) to 4.3 GMACS. We also estimated the environmental impact of the complexity reduction, which can be used as a general guideline for large Transformer-based models, and thus make those models more accessible with less computation overhead.

In this paper, we present a novel mobile user tracking (UT) scheme for codebook-based intelligent reflecting surface (IRS) aided millimeter wave (mmWave) systems. The proposed UT scheme exploits the temporal correlation of the direction from the IRS to the mobile user for selecting IRS phase shifts that provide reflection towards the user. To this end, the user's direction is periodically estimated based on a generalized likelihood ratio test (GLRT) and the user's movement trajectory is extrapolated from several past direction estimates. The efficiency of the proposed UT scheme is evaluated in terms of the average effective rate, which accounts for both the required signaling overhead and the achieved signal to noise ratio (SNR). Our results show that for medium to high SNR, the proposed codebook based UT scheme achieves a higher effective rate than two reference approaches based on full codebook search and optimization of the individual IRS unit cells, respectively., Comment: Has been submitted to 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2023)

Recently, convolutional auto-encoders (CAE) were introduced for image coding. They achieved performance improvements over the state-of-the-art JPEG2000 method. However, these performances were obtained using massive CAEs featuring a large number of parameters and whose training required heavy computational power.\\ In this paper, we address the problem of lossy image compression using a CAE with a small memory footprint and low computational power usage. In order to overcome the computational cost issue, the majority of the literature uses Lagrangian proximal regularization methods, which are time consuming themselves.\\ In this work, we propose a constrained approach and a new structured sparse learning method. We design an algorithm and test it on three constraints: the classical $\ell_1$ constraint, the $\ell_{1,\infty}$ and the new $\ell_{1,1}$ constraint. Experimental results show that the $\ell_{1,1}$ constraint provides the best structured sparsity, resulting in a high reduction of memory and computational cost, with similar rate-distortion performance as with dense networks.

Audio captioning aims at describing acoustic scenes with natural language. Systems are currently evaluated by image cap-tioning metrics CIDEr and SPICE. However, recent studies have highlighted a poor correlation of these metrics with human assessments. In this paper, we propose SPICE+, a modification of SPICE that improves caption annotation and comparison with pre-trained language models. The metric parses captions to semantic graphs with a deep dependency annotation model and a refined set of linguistic rules, then compares sentence embeddings of candidate and reference semantic elements. We formulate a score for general-purpose captioning evaluation, that can be tailored to more specific applications. Combined with fluency error detection, the metric achieves competitive performance on the FENSE benchmark, with 84.0% accuracy on AudioCaps and 74.1% on Clotho.Further experiments show that the metric behaves similarly to the full sentence embedding similarity, while the decomposition into semantic elements allows better interpretability of scores and can provide additional information on the properties of captioning systems.

Although deep learning is the mainstream method in unsupervised anomalous sound detection, Gaussian Mixture Model (GMM) with statistical audio frequency representation as input can achieve comparable results with much lower model complexity and fewer parameters. Existing statistical frequency representations, e.g, the log-Mel spectrogram's average or maximum over time, do not always work well for different machines. This paper presents Time-Weighted Frequency Domain Representation (TWFR) with the GMM method (TWFR-GMM) for anomalous sound detection. The TWFR is a generalized statistical frequency domain representation that can adapt to different machine types, using the global weighted ranking pooling over time-domain. This allows GMM estimator to recognize anomalies, even under domain-shift conditions, as visualized with a Mahalanobis distance-based metric. Experiments on DCASE 2022 Challenge Task2 dataset show that our method has better detection performance than recent deep learning methods. TWFR-GMM is the core of our submission that achieved the 3rd place in DCASE 2022 Challenge Task2., Comment: To appear at ICASSP 2023