Zikai Wang

ORCID Identifier(s)


Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering


Electrical Engineering

First Advisor

Qilian Liang


How to allocate resources in the era of Big Data in telecommunications becomes a new issue. Smartphone data could be a function of personality, as the smartphone supports interpersonal interaction, and the data collected from the smartphone usage often contains rich customer opinion and behavioral information. A bandwidth allocation method based on smartphone users' personality traits and channel condition is studied in a unified mathematical framework in this dissertation. Personalizing bandwidth allocation could be done by analyzing smartphone users' personality traits, resulting in business intelligence, a smarter and more efficient usage of the limited bandwidth, while taking channel fading conditions into account. Using the diagnostic inference, the service provider could calculate user's probability of having each personality trait stand on its data usage. One step further, its bandwidth usage of the following period can be predicted using predictive inference. For our proposed bandwidth allocation scheme, both the outage capacity and outage probability are studied in fading channel. Therefore, service providers shall better allocate the limited bandwidth, provide more personal service to each user, and adjust the bandwidth allocation further on account of the real channel condition. How much information can be transported, i.e., the transmission rate, is another subject of great interest in hybrid wireless networks. The main focus of this paper is the effect of channel fading in hybrid wireless network, in which a wired network of base stations is deployed to support long-range communications between wireless nodes. Two types of transmission mode in hybrid wireless network, i.e., intra-cell mode and infrastructure mode, are considered. To effectively overcome fading impairment, optimal multiple access technique is applied, allowing opportunistic sources to transmit concurrently with the scheduled source. Those different sources, much like in wideband CDMA system, share the entire bandwidth. A successive interference cancellation (SIC) strategy is then introduced at receiver side to limit the intra-cell interference and achieve the maximum capacity. Meanwhile, frequency reuse scheme is employed to minimize the inter-cell interference. Since the outage capacity over different fading channels will exhibit different asymptotic behaviors, in this dissertation we examine the Rayleigh, Rician and Nakagami-m models, which are the most commonly used fading models. Close-form solutions for outage throughput capacity at high signal-to-noise-plus-interference ratio (SNIR) are derived. It is showed that, with opportunistic sources, the intra-cell mode effectively combats fading as wireless nodes increases; however, the infrastructure mode is bottlenecked by the downlink transmission since base station is the only transmitter in the cell during the downlink phase. The theoretical bounds obtained and proofs are instrumental to the future network modeling and design. Array signal processing plays an important role in many areas. Besides the Uniform Linear Array, their are many sparse array that have been proposed, for example, Minimum-redundancy array, Co-prime array, nested array, etc. However, most of the array structures have certain disadvantages. A new type of non-linear sparse sensor array called sparse convolutional array is illustrated which can reduce the number of physical sensors while remain a decent performance in DOA estimation. The sparse convolutional array contains three groups of physical sensors and is able to form a hole-free difference co-array. By adding sensors on two sides instead of the center, the proposed array shows improved performance compares to some other approaches while reminds few physical sensors. The array geometrical structure is illustrated and the numerical result is provided. We also extended this array structure to two-dimension case and the performance is illustrated. Automatic modulation classification (AMC) is a fundamental process in wireless communication. In our work, we proposed a Convolutional neural network based Automatic Modulation Classification (AMC) over Rician Fading Channel. We constructed a system to simulate the received modulated samples through Rician fading channel. 6 commonly used modulation methods are considered, including BPSK, QPSK, 8-PSK, 4-PAM, 16-QAM and 64-QAM. The scheme of proposed convolutional neural network classifier is illustrated, and the neural network is trained with the generated samples and the classification accuracy is demonstrated. The classification accuracy of different mod methods are shown, respectively. And accuracies among different K-factor and maximum Doppler are investigated. Finally, we demonstrated the bit error rate of the system assume a successive modulation detection.


Bandwidth allocation, Channel condition, Personality traits, Big data, Business intelligence, Smartphone usage, Hybrid wireless networks, Infrastructure, Fading channel, Outage throughput capacity, Opportunistic communication, Array signal processing, Sparse arrays, Co-array, DOA estimation, Automatic modulation detection, Convolution neural network, Deep learning


Electrical and Computer Engineering | Engineering


Degree granted by The University of Texas at Arlington