Search results for: sensor noise

Authors: Gökhan Silahtaroğlu

Abstract:

Clustering large populations is an important problem when the data contain noise and different shapes. A good clustering algorithm or approach should be efficient enough to detect clusters sensitively. Besides space complexity, time complexity also gains importance as the size grows. Using hierarchies we developed a new algorithm to split attributes according to the values they have and choosing the dimension for splitting so as to divide the database roughly into equal parts as much as possible. At each node we calculate some certain descriptive statistical features of the data which reside and by pruning we generate the natural clusters with a complexity of O(n).

Keywords: Clustering, tree, split, pruning, entropy, gini.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1556

Authors: M. Bouri, S. Bourennane

Abstract:

In this paper, we propose a novel method for subspace estimation used high resolution method without eigendecomposition where the sample Cross-Spectral Matrix (CSM) is replaced by upper triangular matrix obtained from LU factorization. This novel method decreases the computational complexity. The method relies on a recently published result on Rank-Revealing LU (RRLU) factorization. Simulation results demonstrates that the new algorithm outperform the Householder rank-revealing QR (RRQR) factorization method and the MUSIC in the low Signal to Noise Ratio (SNR) scenarios.

Keywords: Factorization, Localization, Matrix, Signalsubspace.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1360

Authors: M. Ketcham, S. Vongpradhip

Abstract:

In this paper, an innovative watermarking scheme for audio signal based on genetic algorithms (GA) in the discrete wavelet transforms is proposed. It is robust against watermarking attacks, which are commonly employed in literature. In addition, the watermarked image quality is also considered. We employ GA for the optimal localization and intensity of watermark. The watermark detection process can be performed without using the original audio signal. The experimental results demonstrate that watermark is inaudible and robust to many digital signal processing, such as cropping, low pass filter, additive noise.

Keywords: Intelligent Audio Watermarking, GeneticAlgorithm, DWT Domain.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2057

Authors: Y. J. Huang, J. D. Yu, B. W. Hong, C. H. Tai, T. C. Kuo

Abstract:

In this paper, a multi-agent robot system is presented. The system consists of four robots. The developed robots are able to automatically enter and patrol a harmful environment, such as the building infected with virus or the factory with leaking hazardous gas. Further, every robot is able to perform obstacle avoidance and search for the victims. Several operation modes are designed: remote control, obstacle avoidance, automatic searching, and so on.

Keywords: autonomous robot, field programmable gate array, obstacle avoidance, ultrasonic sensor, wireless communication.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1778

Authors: G. Sadashivappa, K. V. S. AnandaBabu

Abstract:

The aim of this paper to characterize a larger set of wavelet functions for implementation in a still image compression system using SPIHT algorithm. This paper discusses important features of wavelet functions and filters used in sub band coding to convert image into wavelet coefficients in MATLAB. Image quality is measured objectively using peak signal to noise ratio (PSNR) and its variation with bit rate (bpp). The effect of different parameters is studied on different wavelet functions. Our results provide a good reference for application designers of wavelet based coder.

Keywords: Wavelet, image compression, sub band, SPIHT, PSNR.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2227

Authors: Mazdak Zamani , Azizah A. Manaf , Rabiah B. Ahmad , Akram M. Zeki , Shahidan Abdullah

Abstract:

In this paper, we present a novel, principled approach to resolve the remained problems of substitution technique of audio steganography. Using the proposed genetic algorithm, message bits are embedded into multiple, vague and higher LSB layers, resulting in increased robustness. The robustness specially would be increased against those intentional attacks which try to reveal the hidden message and also some unintentional attacks like noise addition as well.

Keywords: Artificial Intelligence, Audio Steganography, DataHiding, Genetic Algorithm, Substitution Techniques.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 3117

Authors: Fruzsina Pataki-Bittó, Edit Németh

Abstract:

On the one hand, new technologies and communication tools improve employee productivity and accelerate information and knowledge transfer, while on the other hand, information overload and continuous interruptions make it even harder to concentrate at work. It is a great challenge for companies to find the right balance, while there is also an ongoing demand to recruit and retain the talented employees who are able to adopt the modern work style and effectively use modern communication tools. For this reason, this research does not focus on the objective measures of office interruptions, but aims to find those disruption factors which influence the comfort and job satisfaction of employees, and the way how they feel generally at work. The focus of this research is on how employees feel about the different types of interruptions, which are those they themselves identify as hindering factors, and those they feel as stress factors. By identifying and then reducing these destructive factors, job satisfaction can reach a higher level and employee turnover can be reduced. During the research, we collected information from depth interviews and questionnaires asking about work environment, communication channels used in the workplace, individual communication preferences, factors considered as disruptions, and individual steps taken to avoid interruptions. The questionnaire was completed by 141 office workers from several types of workplaces based in Hungary. Even though 66 respondents are working at Hungarian offices of multinational companies, the research is about the characteristics of the Hungarian labor force. The most important result of the research shows that while more than one third of the respondents consider office noise as a disturbing factor, personal inquiries are welcome and considered useful, even if in such cases the work environment will not be convenient to solve tasks requiring concentration. Analyzing the sizes of the offices, in an open-space environment, the rate of those who consider office noise as a disturbing factor is surprisingly lower than in smaller office rooms. Opinions are more diverse regarding information communication technologies. In addition to the interruption factors affecting the employees' job satisfaction, the research also focuses on the role of the offices in the 21^st century.

Keywords: Information overload, interruption, job satisfaction, office environment, work efficiency.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 981

Authors: Anil Kumar, K. Negrat, A. M. Negrat, Abdelsalam Almarimi

Abstract:

In this paper, we present a robust and secure algorithm for watermarking, the watermark is first transformed into the frequency domain using the discrete wavelet transform (DWT). Then the entire DWT coefficient except the LL (Band) discarded, these coefficients are permuted and encrypted by specific mixing. The encrypted coefficients are inserted into the most significant spectral components of the stego-image using a chaotic system. This technique makes our watermark non-vulnerable to the attack (like compression, and geometric distortion) of an active intruder, or due to noise in the transmission link.

Keywords: Blind source separation (BSS), Chaotic system, Watermarking, DWT.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1531

Authors: József Valyon, Gábor Horváth

Abstract:

In comparison to the original SVM, which involves a quadratic programming task; LS–SVM simplifies the required computation, but unfortunately the sparseness of standard SVM is lost. Another problem is that LS-SVM is only optimal if the training samples are corrupted by Gaussian noise. In Least Squares SVM (LS–SVM), the nonlinear solution is obtained, by first mapping the input vector to a high dimensional kernel space in a nonlinear fashion, where the solution is calculated from a linear equation set. In this paper a geometric view of the kernel space is introduced, which enables us to develop a new formulation to achieve a sparse and robust estimate.

Keywords: Support Vector Machines, Least Squares SupportVector Machines, Regression, Sparse approximation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2063

Authors: N. Ben Si Ali, N. Benalia, N. Zarzouri

Abstract:

Speed sensorless systems are intensively studied during recent years; this is mainly due to their economical benefit and fragility of mechanical sensors and also the difficulty of installing this type of sensor in many applications. These systems suffer from instability problems and sensitivity to parameter mismatch at low speed operation. In this paper an analysis of adaptive observer stability with stator resistance estimation is given.

Keywords: Motor drive, sensorless control, adaptive observer, stator resistance estimation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2214

Authors: Wen-Yen Wu, Tsung-Cheng Hsieh, Ching-Sung Lai

Abstract:

In this paper, we propose a method to extract the road signs. Firstly, the grabbed image is converted into the HSV color space to detect the road signs. Secondly, the morphological operations are used to reduce noise. Finally, extract the road sign using the geometric property. The feature extraction of road sign is done by using the color information. The proposed method has been tested for the real situations. From the experimental results, it is seen that the proposed method can extract the road sign features effectively.

Keywords: Color information, image processing, road sign.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2241

Authors: Guy Rosman, Alexander M. Bronstein, Michael M. Bronstein, Ron Kimmel

Abstract:

We present a new algorithm for nonlinear dimensionality reduction that consistently uses global information, and that enables understanding the intrinsic geometry of non-convex manifolds. Compared to methods that consider only local information, our method appears to be more robust to noise. Unlike most methods that incorporate global information, the proposed approach automatically handles non-convexity of the data manifold. We demonstrate the performance of our algorithm and compare it to state-of-the-art methods on synthetic as well as real data.

Keywords: Dimensionality reduction, manifold learning, multidimensional scaling, geodesic distance, boundary detection.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1455

Authors: M. I. Youssef, A. E. Emam, M. Abd Elghany

Abstract:

In this paper, a residue number arithmetic is used in direct sequence spread spectrum system, this system is evaluated and the bit error probability of this system is compared to that of non residue number system. The effect of channel bandwidth, PN sequences, multipath effect and modulation scheme are studied. A Matlab program is developed to measure the signal-to-noise ratio (SNR), and the bit error probability for the various schemes.

Keywords: Spread Spectrum, Direct sequence, Bit errorprobability and Residue number system.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 3650

Authors: J. P. Dubois

Abstract:

Single side band modulation is a widespread technique in communication with significant impact on communication technologies such as DSL modems and ATSC TV. Its widespread utilization is due to its bandwidth and power saving characteristics. In this paper, we present a new scheme for SSB signal generation which is cost efficient and enjoys superior characteristics in terms of frequency stability, selectivity, and robustness to noise. In the process, we develop novel Hilbert transform properties.

Keywords: Crystal filter, frequency drift, frequency mixing, Hilbert transform, phasing, selectivity, single side band AM.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1412

Authors: O. P. Sharma, V. Janyani, S. Sancheti

Abstract:

Data rate, tolerable bit error rate or frame error rate and range & coverage are the key performance requirement of a communication link. In this paper performance of MFSK link is analyzed in terms of bit error rate, number of errors and total number of data processed. In the communication link model proposed, which is implemented using MATLAB block set, an improvement in BER is observed. Different parameters which effects and enables to keep BER low in M-ary communication system are also identified.

Keywords: Additive White Gaussian Noise (AWGN), Bit Error Rate (BER), Frequency Shift Keying (FSK), Orthogonal Signaling.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2889

Authors: Youngpo Lee, Seokho Yoon

Abstract:

In this paper, we propose a distance estimation scheme for radar systems using direct sequence ultra wideband (DS-UWB) signals. The proposed distance estimation scheme averages out the noise by accumulating the correlator outputs of the radar, and thus, helps the radar to employ a short-length DS-UWB signal reducing the correlation processing time. Numerical results confirm that the proposed distance estimation scheme provides a better estimation performance and a reduced correlation processing time compared with those of the conventional DS-UWB radars.

Keywords: Radar, DS-UWB, distance estimation, correlation accumulation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2022

Authors: B. Cortés, A. Boza, D. Pérez, L. Cuenca

Abstract:

The Internet of Things (IoT) field has been applied in industries with different purposes. Sensing Enterprise (SE) is an attribute of an enterprise or a network that allows it to react to business stimuli originating on the Internet. These fields have come into focus recently on the enterprises, and there is some evidence of the use and implications in supply chain management, while finding it as an interesting aspect to work on. This paper presents a revision and proposals of IoT applications in supply chain management.

Keywords: Internet of Things, Sensing Enterprises, Supply Chain Management, Industrial, Production Systems, Sensor.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 5424

Authors: Rahib H. Abiyev

Abstract:

In this paper the application of neuro-fuzzy system for equalization of channel distortion is considered. The structure and operation algorithm of neuro-fuzzy equalizer are described. The use of neuro-fuzzy equalizer in digital signal transmission allows to decrease training time of parameters and decrease the complexity of the network. The simulation of neuro-fuzzy equalizer is performed. The obtained result satisfies the efficiency of application of neurofuzzy technology in channel equalization.

Keywords: Neuro-fuzzy system, noise equalization, neuro-fuzzy equalizer, neural system.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1632

Authors: Sang-Chul Kim, Sung-Il Chien

Abstract:

The error diffusion method generates worm artifacts, and weakens the edge of the halftone image when the continuous gray scale image is reproduced by a binary image. First, to enhance the edges, we propose the edge-enhancing filter by considering the quantization error information and gradient of the neighboring pixels. Furthermore, to remove worm artifacts often appearing in a halftone image, we add adaptively random noise into the weights of an error filter.

Keywords: Artifact suppression, Edge enhancement, Error diffusion method, Halftone image

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1424

Authors: Adnan Al-Smadi, Mahmoud Smadi

Abstract:

This paper presents an algorithm for reconstructing phase and magnitude responses of the impulse response when only the output data are available. The system is driven by a zero-mean independent identically distributed (i.i.d) non-Gaussian sequence that is not observed. The additive noise is assumed to be Gaussian. This is an important and essential problem in many practical applications of various science and engineering areas such as biomedical, seismic, and speech processing signals. The method is based on evaluating the bicepstrum of the third-order statistics of the observed output data. Simulations results are presented that demonstrate the performance of this method.

Keywords: Cepstrum, bicepstrum, third order statistics

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2037

Authors: Michael Schukat

Abstract:

Electromagnetic flowmeters with DC excitation are used for a wide range of fluid measurement tasks, but are rarely found in dosing applications with short measurement cycles due to the achievable accuracy. This paper will identify a number of factors that influence the accuracy of this sensor type when used for short-term measurements. Based on these results a new signal-processing algorithm will be described that overcomes the identified problems to some extend. This new method allows principally a higher accuracy of electromagnetic flowmeters with DC excitation than traditional methods.

Keywords: Electromagnetic Flowmeter, Kalman Filter, ShortMeasurement Cycles, Signal Estimation

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1614

Authors: Davide D’Aria, Paolo Falcone, Luigi Maggi, Aldo Cero, Giovanni Amoroso

Abstract:

The paper presents a methodology for real-time structural health monitoring and geophysical applications. The key elements of the system are a high performance MIMO RADAR sensor, an optical camera and a dedicated set of software algorithms encompassing interferometry, tomography and photogrammetry. The MIMO Radar sensor proposed in this work, provides an extremely high sensitivity to displacements making the system able to react to tiny deformations (up to tens of microns) with a time scale which spans from milliseconds to hours. The MIMO feature of the system makes the system capable of providing a set of two-dimensional images of the observed scene, each mapped on the azimuth-range directions with noticeably resolution in both the dimensions and with an outstanding repetition rate. The back-scattered energy, which is distributed in the 3D space, is projected on a 2D plane, where each pixel has as coordinates the Line-Of-Sight distance and the cross-range azimuthal angle. At the same time, the high performing processing unit allows to sense the observed scene with remarkable refresh periods (up to milliseconds), thus opening the way for combined static and dynamic structural health monitoring. Thanks to the smart TX/RX antenna array layout, the MIMO data can be processed through a tomographic approach to reconstruct the three-dimensional map of the observed scene. This 3D point cloud is then accurately mapped on a 2D digital optical image through photogrammetric techniques, allowing for easy and straightforward interpretations of the measurements. Once the three-dimensional image is reconstructed, a 'repeat-pass' interferometric approach is exploited to provide the user of the system with high frequency three-dimensional motion/vibration estimation of each point of the reconstructed image. At this stage, the methodology leverages consolidated atmospheric correction algorithms to provide reliable displacement and vibration measurements.

Keywords: Interferometry, MIMO RADAR, SAR, tomography.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 911

Authors: Mario Mastriani, Alberto E. Giraldez

Abstract:

We describe a novel method for removing noise (in wavelet domain) of unknown variance from microarrays. The method is based on a smoothing of the coefficients of the highest subbands. Specifically, we decompose the noisy microarray into wavelet subbands, apply smoothing within each highest subband, and reconstruct a microarray from the modified wavelet coefficients. This process is applied a single time, and exclusively to the first level of decomposition, i.e., in most of the cases, it is not necessary a multirresoltuion analysis. Denoising results compare favorably to the most of methods in use at the moment.

Keywords: Directional smoothing, denoising, edge preservation, microarrays, thresholding, wavelets

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1503

Authors: Ayyaz Mahmood

Abstract:

A new approach for the improvement of coding gain in channel coding using Advanced Encryption Standard (AES) and Maximum A Posteriori (MAP) algorithm is proposed. This new approach uses the avalanche effect of block cipher algorithm AES and soft output values of MAP decoding algorithm. The performance of proposed approach is evaluated in the presence of Additive White Gaussian Noise (AWGN). For the verification of proposed approach, computer simulation results are included.

Keywords: Advanced Encryption Standard (AES), Avalanche Effect, Maximum A Posteriori (MAP), Soft Input Decryption (SID).

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1947

Authors: Hadi Sadoghi Yazdi, Tahereh Royani, Mehri Sadoghi Yazdi, Sohrab Effati

Abstract:

In this paper, a new version of support vector regression (SVR) is presented namely Fuzzy Cost SVR (FCSVR). Individual property of the FCSVR is operation over fuzzy data whereas fuzzy cost (fuzzy margin and fuzzy penalty) are maximized. This idea admits to have uncertainty in the penalty and margin terms jointly. Robustness against noise is shown in the experimental results as a property of the proposed method and superiority relative conventional SVR.

Keywords: Support vector regression, Fuzzy input, Fuzzy cost.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1372

Authors: S. Behnam Malekzadeh, I. Kerr, T. Kaempffer, T. Harper, A Watson

Abstract:

The Site C Hydroelectric dam is currently being constructed in north-eastern British Columbia on sub-horizontal sedimentary strata that dip approximately 15 meters from one bank of the Peace River to the other. More than 615 pressure sensors (Vibrating Wire Piezometers) have been installed on bedding planes (BPs) since construction began, with over 80 more planned before project completion. These pressure measurements are essential to monitor the stability of the rock foundation during and after construction and for dam safety purposes. BPs are identified by their clay gouge infilling, which varies in thickness from less than 1 to 20 mm and can be challenging to identify as the core drilling process often disturbs or washes away the gouge material. Without the use of depth predictions from nearby boreholes, stratigraphic markers, and downhole geophysical data, it is difficult to confidently identify BP targets for the sensors. In this paper, a Case-Based Reasoning (CBR) method was used to develop an empirical model called the Bedding Plane Elevation Prediction (BPEP) to help geologists and geotechnical engineers to predict geological features and BPs at new locations in a fast and accurate manner. To develop CBR, a database was developed based on 64 pressure sensors already installed on key bedding planes BP25, BP28, and BP31 on the Right Bank, including BP elevations and coordinates. 13 (20%) of the most recent cases were selected to validate and evaluate the accuracy of the developed model, while the similarity was defined as the distance between previous cases and recent cases to predict the depth of significant BPs. The average difference between actual BP elevations and predicted elevations for above BPs was ± 55 cm, while the actual results showed that 69% of predicted elevations were within ± 79 cm of actual BP elevations while 100% of predicted elevations for new cases were within ± 99 cm range. Eventually, the actual results will be used to develop the database and improve BPEP to perform as a learning machine to predict more accurate BP elevations for future sensor installations.

Keywords: Case-Based Reasoning, CBR, geological feature, geology, piezometer, pressure sensor, core logging, dam construction.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 228

Authors: Cheima Ben Soltane, Ittansa Yonas Kelbesa

Abstract:

Speaker Identification (SI) is the task of establishing identity of an individual based on his/her voice characteristics. The SI task is typically achieved by two-stage signal processing: training and testing. The training process calculates speaker specific feature parameters from the speech and generates speaker models accordingly. In the testing phase, speech samples from unknown speakers are compared with the models and classified. Even though performance of speaker identification systems has improved due to recent advances in speech processing techniques, there is still need of improvement. In this paper, a Closed-Set Tex-Independent Speaker Identification System (CISI) based on a Multiple Classifier System (MCS) is proposed, using Mel Frequency Cepstrum Coefficient (MFCC) as feature extraction and suitable combination of vector quantization (VQ) and Gaussian Mixture Model (GMM) together with Expectation Maximization algorithm (EM) for speaker modeling. The use of Voice Activity Detector (VAD) with a hybrid approach based on Short Time Energy (STE) and Statistical Modeling of Background Noise in the pre-processing step of the feature extraction yields a better and more robust automatic speaker identification system. Also investigation of Linde-Buzo-Gray (LBG) clustering algorithm for initialization of GMM, for estimating the underlying parameters, in the EM step improved the convergence rate and systems performance. It also uses relative index as confidence measures in case of contradiction in identification process by GMM and VQ as well. Simulation results carried out on voxforge.org speech database using MATLAB highlight the efficacy of the proposed method compared to earlier work.

Keywords: Feature Extraction, Speaker Modeling, Feature Matching, Mel Frequency Cepstrum Coefficient (MFCC), Gaussian mixture model (GMM), Vector Quantization (VQ), Linde-Buzo-Gray (LBG), Expectation Maximization (EM), pre-processing, Voice Activity Detection (VAD), Short Time Energy (STE), Background Noise Statistical Modeling, Closed-Set Tex-Independent Speaker Identification System (CISI).

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1889

Authors: Xin Cheng, Benny Thörnberg, Abdul Waheed Malik, Najeem Lawal

Abstract:

We present a hardware oriented method for real-time measurements of object-s position in video. The targeted application area is light spots used as references for robotic navigation. Different algorithms for dynamic thresholding are explored in combination with component labeling and Center Of Gravity (COG) for highest possible precision versus Signal-to-Noise Ratio (SNR). This method was developed with a low hardware cost in focus having only one convolution operation required for preprocessing of data.

Keywords: Dynamic thresholding, segmentation, position measurement, sub-pixel precision, center of gravity.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2353

Authors: Safdar Tanweer, Abdul Mobin, Afshar Alam

Abstract:

The performance and analysis of speech recognition system is illustrated in this paper. An approach to recognize the English word corresponding to digit (0-9) spoken by 2 different speakers is captured in noise free environment. For feature extraction, speech Mel frequency cepstral coefficients (MFCC) has been used which gives a set of feature vectors from recorded speech samples. Neural network model is used to enhance the recognition performance. Feed forward neural network with back propagation algorithm model is used. However other speech recognition techniques such as HMM, DTW exist. All experiments are carried out on Matlab.

Keywords: Speech Recognition, MFCC, Neural Network, classifier.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 3268

Authors: Thomas Bryan, Veton Kepuska, Ivica Kostanic

Abstract:

Speaker recognition is performed in high Additive White Gaussian Noise (AWGN) environments using principals of Computational Auditory Scene Analysis (CASA). CASA methods often classify sounds from images in the time-frequency (T-F) plane using spectrograms or cochleargrams as the image. In this paper atomic decomposition implemented by matching pursuit performs a transform from time series speech signals to the T-F plane. The atomic decomposition creates a sparsely populated T-F vector in “weight space” where each populated T-F position contains an amplitude weight. The weight space vector along with the atomic dictionary represents a denoised, compressed version of the original signal. The arraignment or of the atomic indices in the T-F vector are used for classification. Unsupervised feature learning implemented by a sparse autoencoder learns a single dictionary of basis features from a collection of envelope samples from all speakers. The approach is demonstrated using pairs of speakers from the TIMIT data set. Pairs of speakers are selected randomly from a single district. Each speak has 10 sentences. Two are used for training and 8 for testing. Atomic index probabilities are created for each training sentence and also for each test sentence. Classification is performed by finding the lowest Euclidean distance between then probabilities from the training sentences and the test sentences. Training is done at a 30dB Signal-to-Noise Ratio (SNR). Testing is performed at SNR’s of 0 dB, 5 dB, 10 dB and 30dB. The algorithm has a baseline classification accuracy of ~93% averaged over 10 pairs of speakers from the TIMIT data set. The baseline accuracy is attributable to short sequences of training and test data as well as the overall simplicity of the classification algorithm. The accuracy is not affected by AWGN and produces ~93% accuracy at 0dB SNR.

Keywords: Time-frequency plane, atomic decomposition, envelope sampling, Gabor atoms, matching pursuit, sparse dictionary learning, sparse autoencoder.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1570