Katrina Neville

Compressed image transmission over FFT-OFDM: A comparative study

In this work we present a comparative study for the transmission of a wavelet-compressed still image using FFT-OFDM over multipath channels with additive white Gaussian noise (AWGN). Rather than using a global threshold for the compression of the wavelet coefficients, we chose to compress the data by transmitting only the low-pass subbands from each decomposition level. 16-QAM is considered for the modulation of the data and various wavelets are considered. A comparison is also performed between the common discrete cosine transform (DCT) compression method and the DWT to compare accuracy and efficiency of data transmission using the two methods. The performance criteria for our compression methods include the compression ratio and relative root-mean-squared (RMS) error of the received data.

Recognition of Modulated Speech over OFDMA

This paper presents an analysis of the error associated with speech after it has been processed using QAM modulation and sent through an IFFT/FFT OFDM transmission system. The main focus will be on how this type of modulation and transmission system will affect the unique features of speech used for speaker recognition and verification systems. Channel deformations can lead to serious errors in speech feature, which would result in recognition and verification algorithms becoming unreliable for secure applications. We will use the well-known mel-frequency cepstral feature extraction algorithm to extract the speaker dependent features before and after the speech is modulated and sent over a noisy communication channel for comparison in our study

Performance of a Text-Independent Remote Speaker Recognition Algorithm over Communication Channels with Blind Equalisation

In this work we will present a study of the reliability of a well-known speaker recognition algorithm when using speech sent over communication channels with channel distortion and noise. The speech features used to test and train this system are the mel-frequency cepstral coefficients. For speaker recognition applications, channel deformations can lead to serious errors in recognition if the speech is transmitted, making the algorithm unreliable for usage in telephone banking or other applications requiring a high level of security. We will study the performance and reliability of this algorithm for text-independent speaker recognition with speech sent over a communication channel. We will be using blind equalisation techniques with QPSK modulation.

Facial expression recognition over FFT-OFDM

Facial expression recognition (FER) is an important task in human- computer interaction systems to include emotion processing. In this work we present the recognition of facial expressions in OFDM systems. The original image is converted to binary code then it is transmitted via OFDM over a multi-path fading channel. The features are then extracted from the reconstructed image at the receiver based on the log Gabor filter. Finally, the features are classified using the naive Bayesian classifier. Simulation results on the Cohn-Kanade database show that the system obtains good results in the presence of noise in the received images.

FFT-OFDM for compressed image transmission: Performance using structural similarity

While root-mean squared error (RMSE) is a good indicator of error in a received image, it does not always take into account the structure of the image and the way images are perceived by the human eye. Using the newly-proposed structural similarity image measure (SSIM), wavelets previously studied by the authors were analysed again. Results showed a close relationship between RMSE and SSIM, and using a combination of both techniques the Daubechies wavelet family gave slightly better quality images than Biorthogonal family. The auto-correlation of the received images was also used to quantify structural loss.

An Interference Cancellation Algorithm for Fourier-based and wavelet-based OFDM systems

An interference cancellation algorithm (ICA) for both Fourier- and wavelet-based OFDM is proposed. In this study we investigated both an ideal case as well as a non-ideal case of our system. In the ideal case we assume the reference signal is an unwanted sinusoidal signal (the input to our ICA algorithm). In the non-ideal case, we assume the received OFDM signal is contaminated with sinusoidal interference. Simulation results performed at a SNR of 5dB resulted in an error measurement of 0.01812 for wavelet-OFDM and an error of approximately 0.03277 for Fourier-OFDM for the ideal case and for the non-ideal case an error of 0.01817 was measured for wavelet-based OFDM and 0.03167 for Fourier-based OFDM. This indicates that the proposed ICA exhibits outstanding performance since the results obtained in both cases are almost the same. The wavelet-based OFDM outperformed the Fourier-based OFDM in both cases.