Hazrat Ali

DWT features performance analysis for automatic speech recognition of Urdu

This paper presents the work on Automatic Speech Recognition of Urdu language, using a comparative analysis for Discrete Wavelets Transform (DWT) based features and Mel Frequency Cepstral Coefficients (MFCC). These features have been extracted for one hundred isolated words of Urdu, each word uttered by ten different speakers. The words have been selected from the most frequently used words of Urdu. A variety of age and dialect has been covered by using a balanced corpus approach. After extraction of features, the classification has been achieved by using Linear Discriminant Analysis. After the classification task, the confusion matrix obtained for the DWT features has been compared with the one obtained for Mel-Frequency Cepstral Coefficients based speech recognition. The framework has been trained and tested for speech data recorded under controlled environments. The experimental results are useful in determination of the optimum features for speech recognition task.

Speaker recognition with hybrid features from a deep belief network

Learning representation from audio data has shown advantages over the handcrafted features such as mel-frequency cepstral coefficients (MFCCs) in many audio applications. In most of the representation learning approaches, the connectionist systems have been used to learn and extract latent features from the fixed length data. In this paper, we propose an approach to combine the learned features and the MFCC features for speaker recognition task, which can be applied to audio scripts of different lengths. In particular, we study the use of features from different levels of deep belief network for quantizing the audio data into vectors of audio word counts. These vectors represent the audio scripts of different lengths that make them easier to train a classifier. We show in the experiment that the audio word count vectors generated from mixture of DBN features at different layers give better performance than the MFCC features. We also can achieve further improvement by combining the audio word count vector and the MFCC features.

Differential based area efficient ROM-less Quadrature Direct Digital Frequency Synthesis

Quadrature Direct Digital Frequency Synthesis (QDDFS) is a technique which can generate sine and cosine values in digital domain. In this paper we present a novel architecture for implementing QDDFS, suitable for implementation in Very Large Scale Integration (VLSI). The algorithm is very simple, which leads to a reduced hardware complexity. It takes two seed values and depending upon the input frequency control word, the algorithm is capable of generating different frequencies. Hardware is ROM-less, making the proposed scheme more area efficient than ROM based lookup table algorithms. In terms of speed performance, the architecture proposed in this paper is more time efficient than COordinate Rotation DIgital Computer (CORDIC) based QDDFS and Singletons Method. Simulation results show that the generated values are very close to the actual values.

Automatic speech recognition of Urdu words using linear discriminant analysis

Urdu is amongst the five largest languages of the world and possess a very important role as it shares its vocabulary with languages as Arabic, Persian, Hindi and several other languages of the Indo-Pak. The Automatic Speech Recognition task of Urdu has not been addressed significantly. This paper presents the statistical based classification technique to achieve the task of Automatic Speech Recognition of isolated words in Urdu. The proposed approach is based on calculation of 52 Mel Frequency Cepstral Coefficients for each isolated word. The classification has been achieved with Linear Discriminant Analysis. The successful or incorrect matches have been presented in the Confusion Matrix. As a prototype, the framework has been trained with audio samples of seven speakers including male/female, native/non-native and speakers with different ages. The test set comprises of audio data of three speaker. For each isolated, percentage error has been calculated. It was found that majority of the words are recognized with percentage error less than 33%. Some words suffer 100% error and were referred to be the bad words. This work may provide a baseline for further research on Urdu Automatic Speech Recognition.

An Overview of Bayesian Network Applications in Uncertain Domains

 Abstract—Uncertainty is a major barrier in knowledge discovery from complex problem domains. Knowledge discovery in such domains requires qualitative rather than quantitative analysis. Therefore, the quantitative measures can be used to represent uncertainty with the integration of various models. The Bayesian Network (BN) is a widely applied technique for characterization and analysis of uncertainty in real world domains. Thus, the real application of BN can be observed in a broad range of domains such as image processing, decision making, system reliability estimation and PPDM (Privacy Preserving in Data Mining) in association rule mining and medical domain analysis. BN techniques can be used in these domains for prediction and decision support. In this article, a discussion on general BN representation, draw inferences, learning and prediction is followed by applications of BN in some specific domains. Domain specific BN representation, inferences and learning process are also presented. Building upon the knowledge presented, some future research directions are also highlighted.

Linear Discriminant Analysis Based Approach for Automatic Speech Recognition of Urdu Isolated Words

Urdu is amongst the five largest languages of the world and enjoys extreme importance by sharing its vocabulary with several other languages of the Indo-Pak. However, there has not been any significant research in the area of Automatic Speech Recognition of Urdu. This paper presents the statistical based classification technique to achieve the task of Automatic Speech Recognition of isolated words in Urdu. For each isolated word, 52 Mel Frequency Cepstral Coefficients have been extracted and based upon these coefficients; the classification has been achieved using Linear Discriminant Analysis. As a prototype, the system has been trained with audio samples of seven speakers including male/female, native/non-native and speakers with different ages while the testing has been done using audio samples of three speakers. It was determined that majority of words exhibit a percentage error of less than 33 %. Words with 100 % error were declared to be bad words. The work reported in this paper may serve as a strong baseline for future research work on Urdu ASR, especially for continuous speech recognition of Urdu.

Bayesian network scores based text localization in scene images

Text localization in scene images is an essential and interesting task to analyze the image contents. In this work, a Bayesian network scores using K2 algorithm in conjunction with the geometric features based effective text localization method with the help of maximally stable extremal regions (MSERs). First, all MSER-based extracted candidate characters are directly compared with an existing text localization method to find text regions. Second, adjacent extracted MSER-based candidate characters are not encompassed into text regions due to strict edges constraint. Therefore, extracted candidate character regions are incorporated into text regions using selection rules. Third, K2 algorithm-based Bayesian networks scores are learned for the complimentary candidate character regions. Bayesian logistic regression classifier is built on the Bayesian network scores by computing the posterior probability of complimentary candidate character region corresponding to non-character candidates. The higher posterior probability of complimentary Candidate character regions are further grouped into words or sentences. Bayesian networks scores based text localization system, named as BayesText, is evaluated on ICDAR 2013 Robust Reading Competition (Challenge 2 Task 2.1: Text Localization) database. Experimental results have established significant competitive performance with the state-of-the-art text detection systems.

FPGA architecture for OFDM Software Defined Radio with an optimized Direct Digital Frequency Synthesizer

Withdrawn.

Classifier comparison for MSER-based text classification in scene images

Text detection in images is an emerging area of interest with a growing motivation to researchers. Various methodologies have been developed to localize text contained in scene images. One main application of localizing scene image text is to produce a real time support to visually impaired persons. To design a real-time support platform for visually impaired persons, classification of textual information, i.e. character and non-character information can provide a baseline for further research. However, the challenge exists in choosing the optimum classifier for this purpose. In this work, first, we used Maximally Stable Extremal Regions (MSERs) to detect character candidates in a scene image; then, we trained several classifiers, i.e., AdaboostM1, Bayesian Logistic Regression, Naïve Bayes, and Bayes Net, to classify MSERs as characters and non-characters; and finally, we compared and analyzed the performances of these classifiers empirically. From experiments, it has been concluded that Bayesian Logistic Regression provides the better accuracy over the other three classifiers. This work argues that MSER based character candidates extraction and Bayesian Logistic Regression based text classification are two prominent and potential techniques in scene text detection.

Improving PAPR reduction for OFDM using hybrid techniques

The Peak-to-Average Power Ratio (PAPR) reduction in Orthogonal Frequency Division Multiplexing (OFDM) system has gained widespread attention during the last decade, leading to the development of several techniques for PAPR reduction. To benefit from these techniques while overcoming their inherent shortcomings, more recent work shows the use of hybrid variants of these techniques. In this paper, we present a brief overview of the hybrid techniques adapted for PAPR reduction. We also comment on the computational complexity and practicability of these approaches. Furthermore the paper covers a variety of hybrid techniques combining simple schemes e.g. clipping, companding and more complex schemes such as partial transmit sequence.