scholarly journals A CNN-based approach to identification of degradations in speech signals

Author(s):  
Yuki Saishu ◽  
Amir Hossein Poorjam ◽  
Mads Græsbøll Christensen

AbstractThe presence of degradations in speech signals, which causes acoustic mismatch between training and operating conditions, deteriorates the performance of many speech-based systems. A variety of enhancement techniques have been developed to compensate the acoustic mismatch in speech-based applications. To apply these signal enhancement techniques, however, it is necessary to know prior information about the presence and the type of degradations in speech signals. In this paper, we propose a new convolutional neural network (CNN)-based approach to automatically identify the major types of degradations commonly encountered in speech-based applications, namely additive noise, nonlinear distortion, and reverberation. In this approach, a set of parallel CNNs, each detecting a certain degradation type, is applied to the log-mel spectrogram of audio signals. Experimental results using two different speech types, namely pathological voice and normal running speech, show the effectiveness of the proposed method in detecting the presence and the type of degradations in speech signals which outperforms the state-of-the-art method. Using the score weighted class activation mapping, we provide a visual analysis of how the network makes decision for identifying different types of degradation in speech signals by highlighting the regions of the log-mel spectrogram which are more influential to the target degradation.

Entropy ◽  
2021 ◽  
Vol 23 (11) ◽  
pp. 1502
Author(s):  
Ben Wilkes ◽  
Igor Vatolkin ◽  
Heinrich Müller

We present a multi-modal genre recognition framework that considers the modalities audio, text, and image by features extracted from audio signals, album cover images, and lyrics of music tracks. In contrast to pure learning of features by a neural network as done in the related work, handcrafted features designed for a respective modality are also integrated, allowing for higher interpretability of created models and further theoretical analysis of the impact of individual features on genre prediction. Genre recognition is performed by binary classification of a music track with respect to each genre based on combinations of elementary features. For feature combination a two-level technique is used, which combines aggregation into fixed-length feature vectors with confidence-based fusion of classification results. Extensive experiments have been conducted for three classifier models (Naïve Bayes, Support Vector Machine, and Random Forest) and numerous feature combinations. The results are presented visually, with data reduction for improved perceptibility achieved by multi-objective analysis and restriction to non-dominated data. Feature- and classifier-related hypotheses are formulated based on the data, and their statistical significance is formally analyzed. The statistical analysis shows that the combination of two modalities almost always leads to a significant increase of performance and the combination of three modalities in several cases.


2021 ◽  
Author(s):  
Amira Abdelrasoul

The low-pressure membrane applications are considered to be the most effective and sustainable methods of addressing environmental problems in treating water and wastewater that meets or exceed stringent environmental standards. Nevertheless, membrane fouling is one of the primary operational concerns that is currently hindering a more widespread application of ultrafiltration (UF) with a variety of contaminants. Membrane fouling leads to higher operating costs, higher energy demand, reduced membrane life time, and increased cleaning frequency. As a consequence, an efficient and well-planned UF process is becoming a necessity for consistent and long-term monetary returns. Examining the source and mechanisms of foulant attachment to the membrane’s surface is critical when it comes to the research of membrane fouling and its potential practical implementation. A mathematical model was developed in this study in order to predict the amount of fouling based on an analysis of particle attachments. This model was developed using both homogeneous and heterogeneous membranes, with a uniform and non-uniform pore sizes for the UF of simulated latex effluent with a wide range of particle size distribution. The objective of this mathematical model was to effectively identify and address the common shortcomings of previous fouling models, and to account for the existing chemical attachments in membrane fouling. The mathematical model resulting from this study was capable of accurately predicting the mass of fouling retained by the membrane and the increase in transmembrane pressure (TMP). In addition, predictive models of fouling attachments were derived and now form an extensive set of mathematical models necessary for the prediction of membrane fouling at a given operating condition, as well as, the various membrane surface charges. Polycarbonate and Polysulfone flat membranes, with pore sizes of 0.05 μm and a molecular weight cut off of 60,000 respectively, were used in the experimental designs under a constant feed flow rate and a cross-flow mode in UF of the simulated latex paint effluent. The TMP estimated from the model agreed with the experimentally measured values at different operating conditions, mostly within 5.0 - 8.0 % error, and up to 13.0% error for the uniform, and non-uniform pore size membranes, respectively. Furthermore, different types of membranes with a variety of molecular weight cut-off (MWCO) values were tested so as to evaluate the accuracy of the models for a generalized application. In addition , a power consumption model, incorporating fouling attachment as well as chemical and physical factors in membrane fouling, was developed in order to ensure accurate prediction and scale-up. Innovative remediation techniques were likewise developed and applied in order to minimize membrane fouling, enhance the membrane performance, and save energy. Fouling remediation methodologies included the pre-treating of the latex effluent, so as to limit its fouling propensity by using different types of surfactants as cationic and anionic, in addition to the pH change. The antifouling properties of the membranes were improved through the implementation of the membrane pH treatment and anionic surfactant treatment. Increasing the ionic strength of latex effluent or enhancing the membrane surface hydrophilicity facilitated a significant increase in the cumulative permeate flux, a substantial decrease in the total mass of fouling, and a noticeable decrease in the specific power consumption.


2021 ◽  
pp. 29-38
Author(s):  
Nabeel Ahsan ◽  
Mahrukh Mehmood ◽  
Asad A. Zaidi

This paper discusses different air management technologies for fuel cell systems. Two different types of compressors are analyzed for Proton-exchange membrane fuel cells (PEMFC). Some important criteria are analyzed thoroughly for the selection of turbo compressor among different types of compressors illustrated with the help of matrix representations. The impacts of various input parameters for Fuel Cell (FC) are also explained thoroughly. Later the numerical modeling of an automobile fuel cell system using a high speed turbo-compressor for air supply is explained. The numerical model incorporates the important input parameters related with air and hydrogen. It also performed energy and mass balances across different components such as pump, fan, heat-exchanger, air compressor and also keeps in consideration the pressure drop across the flow pipes and various mechanical parts. The model is solved to obtain the characteristics of the FC system at different operating conditions. Therefore, it can be concluded that the high speed turbo compressor with a turbo-expander can have significant effects on the overall system power and efficiency.


2021 ◽  
Author(s):  
Amira Abdelrasoul

The low-pressure membrane applications are considered to be the most effective and sustainable methods of addressing environmental problems in treating water and wastewater that meets or exceed stringent environmental standards. Nevertheless, membrane fouling is one of the primary operational concerns that is currently hindering a more widespread application of ultrafiltration (UF) with a variety of contaminants. Membrane fouling leads to higher operating costs, higher energy demand, reduced membrane life time, and increased cleaning frequency. As a consequence, an efficient and well-planned UF process is becoming a necessity for consistent and long-term monetary returns. Examining the source and mechanisms of foulant attachment to the membrane’s surface is critical when it comes to the research of membrane fouling and its potential practical implementation. A mathematical model was developed in this study in order to predict the amount of fouling based on an analysis of particle attachments. This model was developed using both homogeneous and heterogeneous membranes, with a uniform and non-uniform pore sizes for the UF of simulated latex effluent with a wide range of particle size distribution. The objective of this mathematical model was to effectively identify and address the common shortcomings of previous fouling models, and to account for the existing chemical attachments in membrane fouling. The mathematical model resulting from this study was capable of accurately predicting the mass of fouling retained by the membrane and the increase in transmembrane pressure (TMP). In addition, predictive models of fouling attachments were derived and now form an extensive set of mathematical models necessary for the prediction of membrane fouling at a given operating condition, as well as, the various membrane surface charges. Polycarbonate and Polysulfone flat membranes, with pore sizes of 0.05 μm and a molecular weight cut off of 60,000 respectively, were used in the experimental designs under a constant feed flow rate and a cross-flow mode in UF of the simulated latex paint effluent. The TMP estimated from the model agreed with the experimentally measured values at different operating conditions, mostly within 5.0 - 8.0 % error, and up to 13.0% error for the uniform, and non-uniform pore size membranes, respectively. Furthermore, different types of membranes with a variety of molecular weight cut-off (MWCO) values were tested so as to evaluate the accuracy of the models for a generalized application. In addition , a power consumption model, incorporating fouling attachment as well as chemical and physical factors in membrane fouling, was developed in order to ensure accurate prediction and scale-up. Innovative remediation techniques were likewise developed and applied in order to minimize membrane fouling, enhance the membrane performance, and save energy. Fouling remediation methodologies included the pre-treating of the latex effluent, so as to limit its fouling propensity by using different types of surfactants as cationic and anionic, in addition to the pH change. The antifouling properties of the membranes were improved through the implementation of the membrane pH treatment and anionic surfactant treatment. Increasing the ionic strength of latex effluent or enhancing the membrane surface hydrophilicity facilitated a significant increase in the cumulative permeate flux, a substantial decrease in the total mass of fouling, and a noticeable decrease in the specific power consumption.


1988 ◽  
Vol 25 (02) ◽  
pp. 75-104
Author(s):  
William J. Sembler

To a pump manufacturer, marine cargo service represents one of the most demanding applications for which he can design and furnish equipment. In addition to being subjected to the stresses encountered in a shipboard environment, cargo pumps must often perform over a wide range of operating conditions and handle multiple fluids with different viscosities, vapor pressures, specific gravities, temperatures, and material requirements. In this paper the author reviews characteristics of the different types of pumps used for marine cargo service, with an emphasis on the special features that should be incorporated into their design for this rigorous duty. Different types of automatic self-priming/stripping systems available for use with these cargo pumps are also examined. Pump operation is discussed, including the significant impact that system design has on proper pump performance.


2021 ◽  
Vol 16 ◽  
Author(s):  
Anshi Lin ◽  
Wei Kong ◽  
Shuaiqun Wang

Background: Advances in brain imaging and high-throughput genotyping techniques have provided new methods for studying the effects of genetic variation on brain structure and function. Traditionally, a variety of prior information has been added into the multivariate regression method for single nucleotide polymorphisms (SNPs) and quantitative traits (QTs) to improve the accuracy of prediction. In previous studies, brain regions of interest (ROIs) with different types of pathological characteristics (Alzheimer's Disease/Mild Cognitive Impairment/healthy control) can only be randomly dispersed in test cases, greatly limiting the prediction ability of the regression model and failing to obtain optimal global results. Objective: This study proposes a multivariate regression model informed by prior diagnostic information to overcome this limitation. Method: In the prediction model, we first consider traditional prior information and then design a new regularization form to integrate the diagnostic information of different sample ROIs into the model. Results: Experiments demonstrated that this method greatly improves the prediction accuracy of the model compared to other methods and selects a batch of promising pathogenic SNP loci. Conclusion: Taking into account that ROIs with different types of pathological characteristics can be employed as prior information, we propose a new method (Diagnosis-Guided Group Sparse Multitask Learning Method) that improves the ability to predict disease-related quantitative feature sites and select genetic feature factors, applying this model to research on the pathogenesis of Alzheimer's disease.


2009 ◽  
pp. 119-136 ◽  
Author(s):  
Barbara F. Westmoreland

In conclusion, this chapter provides an overview of the different types of normal EEG activity and benign variants that are seen in the EEG. One needs to be aware of the normal variability at different ages and different states of wakefulness, drowsiness, and sleep. Dr. Klass has stated that the “detection and interpretation of the EEG data derived from visual analysis involve matters of judgment and experience, which render clinical EEG an art as much as a science.”5


1977 ◽  
Vol 99 (3) ◽  
pp. 309-314 ◽  
Author(s):  
H. C. Simmons

The paper presents data on the drop-size/volume-fraction distributions of sprays observed with a large number of gas-turbine fuel nozzles of different types including both pressure and air-atomizers, using a range of fuel viscosities, at a variety of operating conditions. The data were obtained by both optical and wax-droplet methods. It is shown that a universal nondimensional correlation can be established for all the fuel nozzles when the drop-size is normalized to the mass median diameter. The correlation enables prediction of the drop-size/volume-fraction distribution for a spray given only the mass median or Sauter mean diameter.


Author(s):  
Olaf Diers ◽  
Denis Schneider ◽  
Melanie Voges ◽  
Peter Weigand ◽  
Christoph Hassa

This contribution is a continuation of ASME-GT2006-90300. While still working at atmospheric pressure, the range of operating conditions was extended to more realistic reduced mass flows to reproduce the engine pressure loss and air preheat up to 700K. The thermoacoustic behaviour of the burner was mapped over that operating range. Two different types of oscillations were observed for flames anchored at the nozzle or lifted from it. Both exhibited a frequency dependence on the Strouhal number for constant reduced mass flows. For a selected operating point with the lifted flame at a preheat temperature of 600K and a reduced mass flow of 0.3kg K0.5/(s bar), the thermoacoustic behaviour of the burner was characterised by phase locked Particle Image Velocimetry as well as phase locked OH- and OH-T- LIF measurements and correlated to the acoustic pressure signal obtained by microphones. The combined data showed pulsating combustion being supported through periodic reignition of the main flame zone by a recirculating volume of hot, OH-rich gas, the cycle time being connected to the observed frequency. The characterization of the preheated operating point was completed with a heat balance investigation quantifying the non-adiabatic combustion conditions of the uncooled combustor.


Energies ◽  
2020 ◽  
Vol 13 (20) ◽  
pp. 5383
Author(s):  
Hafiz Muhammad Uzair Ayub ◽  
Sang Jin Park ◽  
Michael Binns

To help meet the global demand for energy and reduce the use of fossil fuels, alternatives such as the production of syngas from renewable biomass can be considered. This conversion of biomass to syngas is possible through a thermochemical gasification process. To design such gasification systems, model equations can be formulated and solved to predict the quantity and quality of the syngas produced with different operating conditions (temperature, the flow rate of an oxidizing agent, etc.) and with different types of biomass (wood, grass, seeds, food waste, etc.). For the comparison of multiple different types of biomass and optimization to find optimal conditions, simpler models are preferred which can be solved very quickly using modern desktop computers. In this study, a number of different stoichiometric thermodynamic models are compared to determine which are the most appropriate. To correct some of the errors associated with thermodynamic models, correction factors are utilized to modify the equilibrium constants of the methanation and water gas shift reactions, which allows them to better predict the real output composition of the gasification reactors. A number of different models can be obtained using different correction factors, model parameters, and assumptions, and these models are compared and validated against experimental data and modelling studies from the literature.


Sign in / Sign up

Export Citation Format

Share Document