Teddy Surya Gunawan

English digits speech recognition system based on Hidden Markov Models

This paper aims to design and implement English digits speech recognition system using Matlab (GUI). This work was based on the Hidden Markov Model (HMM), which provides a highly reliable way for recognizing speech. The system is able to recognize the speech waveform by translating the speech waveform into a set of feature vectors using Mel Frequency Cepstral Coefficients (MFCC) technique This paper focuses on all English digits from (Zero through Nine), which is based on isolated words structure. Two modules were developed, namely the isolated words speech recognition and the continuous speech recognition. Both modules were tested in both clean and noisy environments and showed a successful recognition rates. In clean environment and isolated words speech recognition module, the multi-speaker mode achieved 99.5% whereas the speaker-independent mode achieved 79.5%. In clean environment and continuous speech recognition module, the multi-speaker mode achieved 72.5% whereas the speaker-independent mode achieved 56.25%. However in noisy environment and isolated words speech recognition module, the multi-speaker mode achieved 88% whereas the speaker-independent mode achieved 67%. In noisy environment and continuous speech recognition module, the multi-speaker mode achieved 82.5% whereas the speaker-independent mode achieved 76.67%. These recognition rates are relatively successful if compared to similar systems.

Properties of digital image watermarking

Digital image watermarking techniques have been developed widely in recent years to maintain the broadcasting media and content authentication, broadcast monitoring, copy control, and many other applications. Therefore, many studies have used digital image watermarking to solve these problem. This paper highlights digital image watermarking. It starts with a basic model of digital image watermarking, it discusses the main requirements and applications. Moreover, it reviews some of the techniques and algorithm used in image watermarking. In addition, digital image watermarking attacks are discussed. Lastly, Watermarking evaluation system is described.

Perceptual speech enhancement exploiting temporal masking properties of human auditory system

The use of simultaneous masking in speech enhancement has shown promise for a range of noise types. In this paper, a new speech enhancement algorithm based on a short-term temporal masking threshold to noise ratio (MNR) is presented. A novel functional model for forward masking based on three parameters is incorporated into a speech enhancement framework based on speech boosting. The performance of the speech enhancement algorithm using the proposed forward masking model was compared with seven other speech enhancement methods over 12 different noise types and four SNRs. Objective evaluation using PESQ revealed that using the proposed forward masking model, the speech enhancement algorithm outperforms the other algorithms by 6-20% depending on the SNR. Moreover, subjective evaluation using 16 listeners confirmed the objective test results.

Wavelet packet based audio coding using temporal masking

Conventional audio coding algorithms do not exploit knowledge of the temporal masking properties of the human auditory system, relying solely on simultaneous masking models. This paper presents wavelet packet based audio coding incorporating both temporal and simultaneous masking models. By applying a novel temporal masking model, following simultaneous masking, combined masking thresholds were calculated more accurately resulting in a bit rate reduction of approximately 25 kbps while preserving perceptual quality. This result was confirmed by semi-formal subjective test on audio signals.

Design of environmental friendly hybrid electric vehicle

Hybrid electric vehicles (HEVs) powered by electric machines and an internal combustion engine (ICE) are a promising mean of reducing emissions and fuel consumption without compromising vehicle functionality and driving performances. This paper presents the design of an environmental friendly hybrid car that feature the gasoline engine and batteries pack. The fuel consumption benefited by hybridization are benchmarked to conventional Gasoline and Diesel fueled vehicles. The relationship between fuel consumption and vehicle weight is investigated for Gasoline as well as for Diesel fueled vehicles. Although the automobile manufacturers have reduced the greenhouse gases such as hydro-carbons, carbon monoxide, carbon dioxide, etc., from the vehicle, they cannot produce a zero-emission vehicle unless they produce an electric vehicle (EV). An electric vehicle is an emission free, environmental friendly vehicle. The proposed design of HEV intelligently gets around the individual problems associated with the gasoline engine and the electric vehicle. It diminishes the production of emissions and the use of fuel. The problem of batteries for the electric vehicle is conquered. An HEV charges itself; it never has to be plugged in. When it does not provide power, the motor can run as a generator to transfer energy from regenerative braking and from the gasoline engine to the batteries.

Design and implementation of automatic number plate recognition on android platform

Android platform has gained popularity in recent years in terms of market share and number of available applications. Android operating system is built on a modified Linux kernel with built-in services such as email, web browser, and map applications. In this paper, automatic number plate recognition (ANPR) was designed and implemented on Android mobile phone platform. First, the graphical user interface (GUI) for capturing image using built-in camera was developed to acquire car plate number in Malaysia. Second, the preprocessing of raw image was done using contrast enhancement, filtering, and straightening. Next, an optical character recognition (OCR) using neural network was utilized to extract texts and numbers. The proposed ANPR algorithm was implemented and simulated using Android SDK on a computer. The preliminary results showed that our system is able to recognize most of the plate characters by almost 88%. Future research includes optimizing the system for mobile phone implementation with limited CPU and memory resources, and geo-tagging of the image using GPS coordinates and online database for various mobile applications.

Speech compression using compressive sensing on a multicore system

Compressive sensing (CS) is a new approach to simultaneous sensing and compression of sparse and compressible signals, i.e. speech signal. Compressive sensing is a new paradigm of acquiring signals, fundamentally different from uniform rate digitization followed by compression, often used for transmission or storage. In this paper, a novel algorithm for speech coding utilizing CS principle is developed. The sparsity of speech signals is exploited using gammatone filterbank and Discrete Cosine Transform (DCT) in which the compressive sensing principle is then applied to the sparse subband signals. All parameters will be optimized using informal listening test and Perceptual Evaluation of Speech Quality (PESQ). In order to further reduce the bit requirement, vector quantization using codebook of the training signals will be added to the system. The performance of overall algorithms will be evaluated based on the processing time and speech quality. Finally, to speed up the process, the proposed algorithm will be implemented in a multicore system, i.e. six cores, using Single Program Multiple Data (SPMD) parallel paradigm.

Evaluation of sparsifying algorithms for speech signals

Sparse representations of signals have been used in many areas of signal and image processing. It has also played an important role in compressive sensing algorithms since it performs well in sparse signals. A sparse representation is one in which small number of coefficients contain large proportion of the energy. Sparsity is important also in speech compression and coding, where the signal can be compressed in pre-processing stages. It leads to efficient and robust methods for compression, detection denoising and signal separation. The objective of this paper is to evaluate several transforms which is used to sparsify the speech signals. Fast Fourier Transform (FFT), Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) will be compared and evaluated based on Gini Index. Sparsity properties and measures will be reviewed in this paper. Finally, sparse applications in speech compression and compressive sensing will be discussed.

On the optimal window shape for genomic signal processing

The segments of DNA molecule, called genes are known to carry useful information in their protein coding regions (exons) and are responsible for protein synthesis. The most popular frequency domain technique for gene identification is by using a sliding window DFT technique to locate the well known period-3 property in DNA sequences. This paper investigates the use of parametric and nonparametric window types with DFT based period-3 detection method to identify the coding and noncoding regions. The HMR195 dataset was selected for the performance comparison utilizing the receiver operating characteristic (ROC) curve and area under ROC (AUC) measure. Results showed that the Bartlett window (nonparametric) and Gaussian window with alpha=2.2401 (parametric) provide the optimal window shape compared to 13 other window types for gene identification of coding and noncoding regions.

A New Forward Masking Model and its Application to Speech Enhancement

This paper presents a new forward masking model, which is applied to speech enhancement. The model develops a novel expression for forward masking, where the parameters are related to the masker level, the delay and the frequency obtained by curve-fitting the psychoacoustic data. This model is then incorporated, in a novel way, into a speech enhancement scheme. Objective measures using PESQ demonstrates that our enhancement scheme, provides significant improvements over four existing speech enhancement methods, when tested with speech signals corrupted by various noises at very low signal to noise ratios. Hence, the new forward masking model provides a greater and more accurate masking threshold calculation that leads to better PESQ scores