With the expansion in wireless communication technology and the introduction of powerful smart-ph... more With the expansion in wireless communication technology and the introduction of powerful smart-phones, users are demanding systems which will allow for ubiquitous computing. A critical requirement is a simpler means of interacting with mobile devices. Instead of struggling with small keypads on smart-phones or a stylus on a PDA it would be much simpler if we could use a more natural and familiar medium of communication, speech. There are currently 3 architectures, Embedded Speech Recognition, Network Speech Recognition (NSR) and Distributed Speech Recognition (DSR), each with their own pros and cons, which aim to incorporate an Automatic Speech Recognition (ASR) system on mobile devices. DSR proposes to be the best solution due to its superior performance in the presence of transmission errors and noisy environments. The main aim of this paper is to give the reader a broad outline of the DSR architecture, but focuses mainly on the front-end system, which literature suggests is the m...
Microphone arrays offer the possibility of hands free speech acquisition. This increases the conv... more Microphone arrays offer the possibility of hands free speech acquisition. This increases the convenience for those using speech technologies as they do not need to hold a microphone in order to interact with a speech system. In addition, a microphone array also has the advantage of potential gains in signal-tonoise ratio in noisy and reverberant environments. In this paper we evaluate the quality of a locally designed four element linear microphone array. The microphone array enhanced speech is evaluated on distortion, noise and speaker identification performance. The reported results show the noise canceling beamformer with post filter as having produced low distortion, high signal-tonoise ratio speech and the best speaker identification rate when compared to other general beamforming techniques.
Performance of speech recognition systems on speech that has been transmitted through GSM and Tel... more Performance of speech recognition systems on speech that has been transmitted through GSM and Telephone channels is generally very poor. The poor performance in recognition is mainly due to channel effects, which puts limitations on the use of speech recognition applications over telecommunication networks. In an effort to reduce the degradation in speech recognition performance, Speaker normalization and channel normalization, which are two strategies to tackle the variation from speaker, channel and environments are investigated. In this paper two techniques are examined: vocal tract length normalization (VTLN) for speaker normalization and cepstral mean normalization (CMN) for channel normalization. In addition a combination of VTLN and CMN was implemented to account for both the channel effects and variation in vocal tract length effects. Experiments showed that applying speaker normalization and channel normalization in speech recognition systems leads to relative reduction in ...
State of the art speaker identification systems use the Gaussian mixture models (GMM) classifier.... more State of the art speaker identification systems use the Gaussian mixture models (GMM) classifier. Support vector machines (SVM) offers a competing classification algorithm. Both classification methods have been evaluated on speaker recognition tasks and have shown to produce uncorrelated errors with sometimes similar performance. In this paper their performance is compared on different parametric feature-sets, in particular on their response to spectral compression in the feature-sets. It was found that both classifiers respond to spectral compression, with the SVM performance levelling off at higher compressions. Even though for the limited dataset used SVM performance was better than that of GMM, the SVM required several orders of magnitude more in computation time as compared to GMM.
Abstract—Microphone array systems have been an area of active research for several years. The pot... more Abstract—Microphone array systems have been an area of active research for several years. The potential for high quality hands-free speech acquisition in noisy and reflecting environments makes microphone arrays an attractive alternative to conventional close-talking microphones. The signal-enhancement and sourcelocation capabilities of microphone arrays make them applicable to a variety of tasks including teleconferencing, speaker tracking, speaker recognition and speech recognition. In this paper we evaluate techniques for setting up microphone arrays for speaker identification. We propose the use of an active noise canceling beamformer based on the generalized sidelobe canceller (GSC) beamformer. Significant improvements in identification rate are achieved using this method compared to other beamforming techniques investigated in this paper
The last three decades have witnessed an increased concern in data breaches. This problem is more... more The last three decades have witnessed an increased concern in data breaches. This problem is more pronounced in patients' health information (PHI) management. Previous researchers have reported the need to address this problem from a multi-criteria perspective; however, there is sparse information on how to use fuzzy logic to integrated PHI data breach criteria. This article, therefore, presents a fuzzy-based model, which is a fuzzy cognitive map, for a data breach in health-centres using organizational and human criteria. This model combines fuzzy logic, decision-makers' opinions and cognitive maps method. Decision-makers' opinions were analysis from optimistic, pessimistic and most likely perspectives. The performance of this model has evaluated three FCM data breach forecasting scenarios. And the results obtained showed that the model predicts PHI data breach under different multi-criteria conditions. This article has been able to shed insights into the application of a fuzzy cognitive map for data breach problem in health-care centres
2019 IEEE 2nd International Conference on Renewable Energy and Power Engineering (REPE), 2019
The growing needs for clean energy will continue to attract global attention, especially as it ha... more The growing needs for clean energy will continue to attract global attention, especially as it has been recognized as a means of managing solid wastes - especially from households and industrial sector. We now have different waste-to-energy technologies for small and medium-scale plants. But sparse information exists on how to optimize these plants operational efficiencies, especially boilers and reformers. Hence, this article considers the optimization of these efficiencies to optimal electricity generation. This objective is achieved using a nonlinear programming approach. The proposed model utility was tested using a case study of six locations in Southwest Nigeria. A comparison of Genetic algorithm (GA) and Differential Evolution (DE) algorithm are presented as solution methods for the model. In terms of the total electricity generated, there is no significant difference between these algorithms results. The total electricity generated is 10MW, while the average boilers and reformers efficiencies are 0.9 and 0.8, respectively. To be strategic with a waste-to-energy operation, this article recommends that optimal parametric settings for a plant’s operational efficiencies should be combined with experts’ opinions.
Pattern recognition and classification are fundamental concepts for understanding intelligent liv... more Pattern recognition and classification are fundamental concepts for understanding intelligent living systems and essential for realising artificial intelligent system. The field of Pattern Recognition has expanded in recent years to find application in many fields ranging from Image Processing, Language Processing, Data Mining and Retrieval which includes as a special case Content-based Image Retrieval and Non-linear Modelling and Synthesis to name just a few. The scope of this issue will range over all novel applications of Pattern Recognition to solve real life problems as well as novel theoretical contributions to Pattern Recognition Theory. Topics of interest include, but are not limited to, the following:
Microphone array systems have been an area of active research for several years. The potential fo... more Microphone array systems have been an area of active research for several years. The potential for high quality hands-free speech acquisition in noisy and reflecting environments makes microphone arrays an attractive alternative to conventional close-talking microphones. The signal-enhancement and sourcelocation capabilities of microphone arrays make them applicable to a variety of tasks including teleconferencing, speaker tracking, speaker recognition and speech recognition. In this paper we evaluate techniques for setting up microphone arrays for speaker identification. We propose the use of an active noise canceling beamformer based on the generalized sidelobe canceller (GSC) beamformer. Significant improvements in identification rate are achieved using this method compared to other beamforming techniques investigated in this paper.
Following on the development of several prototypes, we have built a semi-automated Deaf Telephony... more Following on the development of several prototypes, we have built a semi-automated Deaf Telephony prototype on the SoftBridge platform. This prototype relays text and speech between Deaf users on the Internet and hearing users on the telephone system. Previous work with a pilot trial in the laboratory revealed several opportunities for enhancement. We added a Wizard of Oz (WoOz) to replace the poorly performing automatic speech recognition functionality as well as H.323 breakout, more extensive logging and advanced call initiation functionality. In order to trial the current prototype, we initiated an Information and Communication Technology (ICT) training programme with the Deaf Community of Cape Town. Twenty Deaf users participated in the training. In addition to the training, much baseline user data was collected to give an indication of how Deaf users communicate with hearing users as well as how familiar they are with ICT devices and services. The work for the rest of this year...
Several studies have called the attentions of utility firms to the possibility of using mathemati... more Several studies have called the attentions of utility firms to the possibility of using mathematical models to measure and monitor energy theft. Unfortunately, these studies have decoupled the contributions of government policies, such as social, technical and economic policies, from their evaluation process. To address this knowledge gaps, this article modelled energy theft using soft computing approach: fuzzy cognitive map (FCM) and swarm algorithm. Fuzzy logic was used to design cognitive maps for energy theft parameters; second, and swarm algorithm was used to determine the weights and concepts values. The practicality of the swarm-based model was tested using experts' judgements. This model performance was compared with evolutionary-based FCM and it was observed that it performed better than the evolutionary-based model. And when the swarm-based model performance was compared with experts' judgements, it performed satisfactorily.
In helping people to access information the use of speech technology presents a very attractive a... more In helping people to access information the use of speech technology presents a very attractive alternative to other methods. In South Africa, there are many people who cannot access computer information due to language and educational level, yet over 15 million of these people own or use cellphones. This paper discusses the work we are doing to enable these users to access information using their speech. We looked at three ways in which the system can be implemented and chose an implementation that will have fewer requirements in terms of the cellphones users can use. We used the HTK toolkit to experiment with system and found that the idea is practical. The next step would be to design the complete system.
Journal of Digital Food, Energy & Water Systems, 2020
Flood management is a global problem that has created immense contributions from researchers and ... more Flood management is a global problem that has created immense contributions from researchers and practitioners, especially those in developing countries. These people often seek ways to minimise the aftermath of a flood. Recently, they are making a case for sustainable solutions to flood management. This study, therefore, contributes a sustainability model that addresses the problem of blue-green technology selection to the current discussion on flood management. It coupled the techno-economic, social, and environmental impact of a blue-green technology using the unique attributions of three multi-criteria decision-making tools: best-worst method, fuzzy axiomatic method and VIKOR; its performance was investigated with qualitative data sets that were obtained from experts. The outcomes of the investigation showed that techno-economic criteria contributed about 88.18% to the ranking of blue-green technology. The most and least suitable blue-green technologies for a community in Nigeri...
Abstract Renewable energy systems (RES) penetration has improved interest in energy storage syste... more Abstract Renewable energy systems (RES) penetration has improved interest in energy storage systems. This has helped to increase RES acceptance because their reliability has improved. Currently, scholars have recognized compressed air energy storage (CAES) systems as efficient storage systems for RES, but there is a need to subject the implications of these systems’ parametric settings on their storage efficiency. Hence, this study proposes and selects a suitable support vector regression model for storage efficiency prediction. It considered five input parameters — among which are Maximum exit temperature and thermal energy. Data sets for CAES with high temperatures (HTE) were used to analyze the performance of different kernels – radial basis function (RBF), polynomial, and linear – for SVR models. These data were divided into train and test examples. The study observed that an RBF trained-SVR model can accurately predict the storage efficiency of CAES-THE, whereas a polynomial trained-SVR cannot predict the storage efficiency of CAES-HTE. The RBF trained-SVR model testing examples correlation coefficient was higher than its training examples result (0.9855) by 1.43%.
IOP Conference Series: Materials Science and Engineering, 2021
Across the world, machining learning (ML) algorithms, such as support vector machines (SVR) and a... more Across the world, machining learning (ML) algorithms, such as support vector machines (SVR) and artificial neural networks (ANN), are among scientific tools for the fourth industrial revolution, or Industry 4.0, campaign. These algorithms have wide engineering applications, but their potentials in energy management are still evolving. Hence, this study investigates the performance of SVR and ANN algorithms as predictive models for wind turbines capacity factor (CF) estimation. Five independent parameters-wind speed, power density, turbulence intensity, installed capacity, and wind shear - were used as input parameters for this estimation problem. Polynomial, radial basis function (RBF), and linear kernels were used to train an SVR model that estimates CF, while Adams was used to optimize the performance of a backpropagation ANN model. These models’ applicability was evaluated using data sets from eight locations. This study used correlation coefficient used to compared the model per...
With the expansion in wireless communication technology and the introduction of powerful smart-ph... more With the expansion in wireless communication technology and the introduction of powerful smart-phones, users are demanding systems which will allow for ubiquitous computing. A critical requirement is a simpler means of interacting with mobile devices. Instead of struggling with small keypads on smart-phones or a stylus on a PDA it would be much simpler if we could use a more natural and familiar medium of communication, speech. There are currently 3 architectures, Embedded Speech Recognition, Network Speech Recognition (NSR) and Distributed Speech Recognition (DSR), each with their own pros and cons, which aim to incorporate an Automatic Speech Recognition (ASR) system on mobile devices. DSR proposes to be the best solution due to its superior performance in the presence of transmission errors and noisy environments. The main aim of this paper is to give the reader a broad outline of the DSR architecture, but focuses mainly on the front-end system, which literature suggests is the m...
Microphone arrays offer the possibility of hands free speech acquisition. This increases the conv... more Microphone arrays offer the possibility of hands free speech acquisition. This increases the convenience for those using speech technologies as they do not need to hold a microphone in order to interact with a speech system. In addition, a microphone array also has the advantage of potential gains in signal-tonoise ratio in noisy and reverberant environments. In this paper we evaluate the quality of a locally designed four element linear microphone array. The microphone array enhanced speech is evaluated on distortion, noise and speaker identification performance. The reported results show the noise canceling beamformer with post filter as having produced low distortion, high signal-tonoise ratio speech and the best speaker identification rate when compared to other general beamforming techniques.
Performance of speech recognition systems on speech that has been transmitted through GSM and Tel... more Performance of speech recognition systems on speech that has been transmitted through GSM and Telephone channels is generally very poor. The poor performance in recognition is mainly due to channel effects, which puts limitations on the use of speech recognition applications over telecommunication networks. In an effort to reduce the degradation in speech recognition performance, Speaker normalization and channel normalization, which are two strategies to tackle the variation from speaker, channel and environments are investigated. In this paper two techniques are examined: vocal tract length normalization (VTLN) for speaker normalization and cepstral mean normalization (CMN) for channel normalization. In addition a combination of VTLN and CMN was implemented to account for both the channel effects and variation in vocal tract length effects. Experiments showed that applying speaker normalization and channel normalization in speech recognition systems leads to relative reduction in ...
State of the art speaker identification systems use the Gaussian mixture models (GMM) classifier.... more State of the art speaker identification systems use the Gaussian mixture models (GMM) classifier. Support vector machines (SVM) offers a competing classification algorithm. Both classification methods have been evaluated on speaker recognition tasks and have shown to produce uncorrelated errors with sometimes similar performance. In this paper their performance is compared on different parametric feature-sets, in particular on their response to spectral compression in the feature-sets. It was found that both classifiers respond to spectral compression, with the SVM performance levelling off at higher compressions. Even though for the limited dataset used SVM performance was better than that of GMM, the SVM required several orders of magnitude more in computation time as compared to GMM.
Abstract—Microphone array systems have been an area of active research for several years. The pot... more Abstract—Microphone array systems have been an area of active research for several years. The potential for high quality hands-free speech acquisition in noisy and reflecting environments makes microphone arrays an attractive alternative to conventional close-talking microphones. The signal-enhancement and sourcelocation capabilities of microphone arrays make them applicable to a variety of tasks including teleconferencing, speaker tracking, speaker recognition and speech recognition. In this paper we evaluate techniques for setting up microphone arrays for speaker identification. We propose the use of an active noise canceling beamformer based on the generalized sidelobe canceller (GSC) beamformer. Significant improvements in identification rate are achieved using this method compared to other beamforming techniques investigated in this paper
The last three decades have witnessed an increased concern in data breaches. This problem is more... more The last three decades have witnessed an increased concern in data breaches. This problem is more pronounced in patients' health information (PHI) management. Previous researchers have reported the need to address this problem from a multi-criteria perspective; however, there is sparse information on how to use fuzzy logic to integrated PHI data breach criteria. This article, therefore, presents a fuzzy-based model, which is a fuzzy cognitive map, for a data breach in health-centres using organizational and human criteria. This model combines fuzzy logic, decision-makers' opinions and cognitive maps method. Decision-makers' opinions were analysis from optimistic, pessimistic and most likely perspectives. The performance of this model has evaluated three FCM data breach forecasting scenarios. And the results obtained showed that the model predicts PHI data breach under different multi-criteria conditions. This article has been able to shed insights into the application of a fuzzy cognitive map for data breach problem in health-care centres
2019 IEEE 2nd International Conference on Renewable Energy and Power Engineering (REPE), 2019
The growing needs for clean energy will continue to attract global attention, especially as it ha... more The growing needs for clean energy will continue to attract global attention, especially as it has been recognized as a means of managing solid wastes - especially from households and industrial sector. We now have different waste-to-energy technologies for small and medium-scale plants. But sparse information exists on how to optimize these plants operational efficiencies, especially boilers and reformers. Hence, this article considers the optimization of these efficiencies to optimal electricity generation. This objective is achieved using a nonlinear programming approach. The proposed model utility was tested using a case study of six locations in Southwest Nigeria. A comparison of Genetic algorithm (GA) and Differential Evolution (DE) algorithm are presented as solution methods for the model. In terms of the total electricity generated, there is no significant difference between these algorithms results. The total electricity generated is 10MW, while the average boilers and reformers efficiencies are 0.9 and 0.8, respectively. To be strategic with a waste-to-energy operation, this article recommends that optimal parametric settings for a plant’s operational efficiencies should be combined with experts’ opinions.
Pattern recognition and classification are fundamental concepts for understanding intelligent liv... more Pattern recognition and classification are fundamental concepts for understanding intelligent living systems and essential for realising artificial intelligent system. The field of Pattern Recognition has expanded in recent years to find application in many fields ranging from Image Processing, Language Processing, Data Mining and Retrieval which includes as a special case Content-based Image Retrieval and Non-linear Modelling and Synthesis to name just a few. The scope of this issue will range over all novel applications of Pattern Recognition to solve real life problems as well as novel theoretical contributions to Pattern Recognition Theory. Topics of interest include, but are not limited to, the following:
Microphone array systems have been an area of active research for several years. The potential fo... more Microphone array systems have been an area of active research for several years. The potential for high quality hands-free speech acquisition in noisy and reflecting environments makes microphone arrays an attractive alternative to conventional close-talking microphones. The signal-enhancement and sourcelocation capabilities of microphone arrays make them applicable to a variety of tasks including teleconferencing, speaker tracking, speaker recognition and speech recognition. In this paper we evaluate techniques for setting up microphone arrays for speaker identification. We propose the use of an active noise canceling beamformer based on the generalized sidelobe canceller (GSC) beamformer. Significant improvements in identification rate are achieved using this method compared to other beamforming techniques investigated in this paper.
Following on the development of several prototypes, we have built a semi-automated Deaf Telephony... more Following on the development of several prototypes, we have built a semi-automated Deaf Telephony prototype on the SoftBridge platform. This prototype relays text and speech between Deaf users on the Internet and hearing users on the telephone system. Previous work with a pilot trial in the laboratory revealed several opportunities for enhancement. We added a Wizard of Oz (WoOz) to replace the poorly performing automatic speech recognition functionality as well as H.323 breakout, more extensive logging and advanced call initiation functionality. In order to trial the current prototype, we initiated an Information and Communication Technology (ICT) training programme with the Deaf Community of Cape Town. Twenty Deaf users participated in the training. In addition to the training, much baseline user data was collected to give an indication of how Deaf users communicate with hearing users as well as how familiar they are with ICT devices and services. The work for the rest of this year...
Several studies have called the attentions of utility firms to the possibility of using mathemati... more Several studies have called the attentions of utility firms to the possibility of using mathematical models to measure and monitor energy theft. Unfortunately, these studies have decoupled the contributions of government policies, such as social, technical and economic policies, from their evaluation process. To address this knowledge gaps, this article modelled energy theft using soft computing approach: fuzzy cognitive map (FCM) and swarm algorithm. Fuzzy logic was used to design cognitive maps for energy theft parameters; second, and swarm algorithm was used to determine the weights and concepts values. The practicality of the swarm-based model was tested using experts' judgements. This model performance was compared with evolutionary-based FCM and it was observed that it performed better than the evolutionary-based model. And when the swarm-based model performance was compared with experts' judgements, it performed satisfactorily.
In helping people to access information the use of speech technology presents a very attractive a... more In helping people to access information the use of speech technology presents a very attractive alternative to other methods. In South Africa, there are many people who cannot access computer information due to language and educational level, yet over 15 million of these people own or use cellphones. This paper discusses the work we are doing to enable these users to access information using their speech. We looked at three ways in which the system can be implemented and chose an implementation that will have fewer requirements in terms of the cellphones users can use. We used the HTK toolkit to experiment with system and found that the idea is practical. The next step would be to design the complete system.
Journal of Digital Food, Energy & Water Systems, 2020
Flood management is a global problem that has created immense contributions from researchers and ... more Flood management is a global problem that has created immense contributions from researchers and practitioners, especially those in developing countries. These people often seek ways to minimise the aftermath of a flood. Recently, they are making a case for sustainable solutions to flood management. This study, therefore, contributes a sustainability model that addresses the problem of blue-green technology selection to the current discussion on flood management. It coupled the techno-economic, social, and environmental impact of a blue-green technology using the unique attributions of three multi-criteria decision-making tools: best-worst method, fuzzy axiomatic method and VIKOR; its performance was investigated with qualitative data sets that were obtained from experts. The outcomes of the investigation showed that techno-economic criteria contributed about 88.18% to the ranking of blue-green technology. The most and least suitable blue-green technologies for a community in Nigeri...
Abstract Renewable energy systems (RES) penetration has improved interest in energy storage syste... more Abstract Renewable energy systems (RES) penetration has improved interest in energy storage systems. This has helped to increase RES acceptance because their reliability has improved. Currently, scholars have recognized compressed air energy storage (CAES) systems as efficient storage systems for RES, but there is a need to subject the implications of these systems’ parametric settings on their storage efficiency. Hence, this study proposes and selects a suitable support vector regression model for storage efficiency prediction. It considered five input parameters — among which are Maximum exit temperature and thermal energy. Data sets for CAES with high temperatures (HTE) were used to analyze the performance of different kernels – radial basis function (RBF), polynomial, and linear – for SVR models. These data were divided into train and test examples. The study observed that an RBF trained-SVR model can accurately predict the storage efficiency of CAES-THE, whereas a polynomial trained-SVR cannot predict the storage efficiency of CAES-HTE. The RBF trained-SVR model testing examples correlation coefficient was higher than its training examples result (0.9855) by 1.43%.
IOP Conference Series: Materials Science and Engineering, 2021
Across the world, machining learning (ML) algorithms, such as support vector machines (SVR) and a... more Across the world, machining learning (ML) algorithms, such as support vector machines (SVR) and artificial neural networks (ANN), are among scientific tools for the fourth industrial revolution, or Industry 4.0, campaign. These algorithms have wide engineering applications, but their potentials in energy management are still evolving. Hence, this study investigates the performance of SVR and ANN algorithms as predictive models for wind turbines capacity factor (CF) estimation. Five independent parameters-wind speed, power density, turbulence intensity, installed capacity, and wind shear - were used as input parameters for this estimation problem. Polynomial, radial basis function (RBF), and linear kernels were used to train an SVR model that estimates CF, while Adams was used to optimize the performance of a backpropagation ANN model. These models’ applicability was evaluated using data sets from eight locations. This study used correlation coefficient used to compared the model per...
Uploads
Papers by Daniel Mashao