PhD Oral Exam - Hossein Kourkchi , Electrical and Computer Engineering
When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.
Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.
In order to have robust video transmission over error prone telecommunication channels several mechanisms are introduced. These mechanisms try to detect, correct or conceal the errors in the received video stream.
In this thesis, the performance of the video codec is improved in terms of error rates without increasing bit rate. This is done by exploiting the residual syntactic/semantic redundancy inside compressed video along with optimizing the configuration of the state-of-the art entropy coding, i.e., binary arithmetic coding, and optimizing the quantization of the channel output. The thesis is divided into four phases.
In the first phase, a breadth-first suboptimal sequential maximum a posteriori (MAP) decoder is employed for joint source-channel arithmetic decoding of H.264 symbols. The proposed decoder uses not only the intentional redundancy inserted via a forbidden symbol (FS) but also exploits residual redundancy by a syntax checker. In contrast to previous methods this is done as each channel bit is decoded. Simulations using intra prediction modes show improvements in error rates, e.g., syntax element error rate reduction by an order of magnitude for channel SNR of 7.33dB. The cost of this improvement is more computational complexity spent on the syntax checking.
In the second phase, the configuration of the FS in the symbol set is studied. The delay probability function, i.e., the probability of the number of bits required to detect an error, is calculated for various FS configurations. The probability of missed error detection is calculated as a figure of merit for optimizing the FS configuration. The simulation results show the effectiveness of the proposed figure of merit, and support the FS configuration in which the FS lies entirely between the other information carrying symbols to be the best.
In the third phase, a new method for estimating the a priori probability of particular syntax elements is proposed. This estimation is based on the interdependency among the syntax elements that were previously decoded. This estimation is categorized as either reliable or unreliable. The decoder uses this prior information when they are reliable, otherwise the MAP decoder considers that the syntax elements are equiprobable and in turn uses maximum likelihood (ML) decoding. The reliability detection is carried out using a threshold on the local entropy of syntax elements in the neighboring macroblocks.
In the last phase, a new measure to assess performance of the channel quantizer is proposed. This measure is based on the statistics of the rank of true candidate among the sorted list of candidates in the MAP decoder. Simulation results shows that a quantizer designed based on the proposed measure is superior to the quantizers designed based on maximum mutual information and minimum mean square error.