博士 / 國立交通大學 / 資訊科學與工程研究所 / 98 / Perceptual audio coding achieves a high compression ratio by exploiting perceptual irrelevance and data redundancy. By using advanced and sophisticated signal processing methods, perceptual audio coding has generated artifacts that are quite different from the traditional distortions. A new audio technology becomes mature through the successful modeling, measuring and control on the artifacts incurred from the technology. With the advance of new coding modules in the state-of-the-art coding methods such as Advanced Audio Coding (AAC), Spectral Band Replication (SBR), and parametric coding, the incurred artifacts are far more difficult to model, measure and control than those caused by previous encoding systems like pulse code modulation. In this dissertation, we take into consideration the MPEG audio, including MP3, AAC, SBR and PS (Parametric Stereo) coding, to explore the compression artifacts from the novel coding methods in terms of principle, generation sources, perception, and related relief methods. We model the audible artifacts through the time-frequency diagrams; consider the artifacts-susceptible music types; analyze the critical encoding technologies incurring these artifacts; and provide empirical verifications for the artifacts. Specifically, we propose an audio patch method for reducing the two zero-quantization artifacts and the fast odd-radix algorithm for computing the type-IV discrete cosine transform in the filterbank computation for breaking the tradeoff of parallelism and numerical distortion in the existing methods. We establish the compact forms for the Temporal Noise Shaping (TNS) in AAC and consider the known artifact named the time-domain aliasing noise. New kinds of artifacts are explored for SBR and PS. We also demonstrate the predictive bias of the linear prediction used in SBR.

博士 / 國立交通大學 / 電子研究所 / 98 / The patterned ground shield (PGS) must be well designed; otherwise they may not at all able to improve the quality factor. Investigations into different strip length, strip spacing and metal layer positions of the slot-type floating shields for wavelength, attenuation loss, and characteristic impedance, which have not yet been conducted before, are performed in this work. In general, the assumption for lumped-equivalent-circuit-model-based techniques is valid only if the lengths of the DUT devices are much smaller than the distances between two ports. However, this is not always true for larger DUT devices and may result in over de-embedding when intrinsic device performance is involved. Therefore, the proposed de-embedding technique can address the problem of over de-embedding. The contribution of the interconnection and the via stack becomes important as the frequencies increase. Unfortunately, currently existing techniques do not account for via stack parasitic contributions. In this dissertation, high-performance transmission lines and improved de-embedding techniques are presented. The slow-wave concept has been used in order to design high-performance transmission lines and reduce the size of the transmission lines. Accurate models that describe the behavior of RF devices are critical for the circuit designs, and improved parasitic de-embedding techniques are proposed as to achieve accurate device characterization. A novel slow-wave transmission line with optimized slot-type floating shields in advanced CMOS technology is presented. Periodical slot-type floating shields are inserted beneath the transmission line to provide the substrate shield and shorten the electromagnetic propagation wavelength. This is the first study that demonstrates how the wavelength, attenuation loss, and characteristic impedance can be adjusted by changing the strip length, the strip spacing, and the metal layer positions of the slot-type floating shields. Wavelength shortening needs to be achieved with a trade-off between the slow-wave effect and the attenuation loss. The slot-type floating shields with different strip lengths, strip spacings and metal layer positions are analyzed. It is concluded that the minimum strip length provides the most optimal result. A design guideline can be established that enables circuit designers to achieve the most appropriate slot-type floating shields for optimal circuit performance. Transmission line test structures were fabricated by using 45 nm CMOS process technology. Both measurement and electro-magnetic (EM) wave simulation were performed up to 50 GHz. Transmission lines are frequently used at a length of half- or quarter-wavelength. With a shortened wavelength, a saving in silicon area of more than 67% can be achieved by using optimized slot-type floating shields. Experimental results demonstrated a higher effective relative permittivity value, improved by a factor of more than 9, and a better quality factor, improved by a factor of more than 6, as compared to conventional transmission lines. A novel transmission line de-embedding technique is presented. With this technique, the left- and right-side ground-signal-ground (GSG) probe pads can be extracted directly using two transmission line test structures of length L and 2L. An additional through structure is designed using via stack de-embedding, which is unique amongst current de-embedding methods. The advantages of the proposed method include the following: (1) a smaller silicon area; (2) the consideration for discontinuity between the pad and interconnect; (3) the consideration for substrate coupling and contact effects; (4) the employment of via stack de-embedding; and (5) the solution to the over de-embedding. The proposed novel methodology could be considered as a breakthrough in the area of ultra-high frequency de-embedding and should enable more accurate RF models to be developed. In the proposed methodology, intrinsic slow-wave coplanar waveguide (CPW) transmission line structures are placed on the inter-level metallization layers, as they are the most appropriate RF device for a cascade-based de-embedding method involving the via stack de-embedding technique. Experimental results have demonstrated that attenuation loss and wavelength can be optimized by changing the metal density and the metal layer positions of the floating shields. With a shortened wavelength, a reduction in silicon area of more than 66% can be achieved by using optimized slot-type floating shields located both above and below the CPW structure.

碩士 / 國立交通大學 / 電信工程研究所 / 98 / Optical and wireless hybrid access network is an integrated access network. Compared with optical access network, it can solve the “last mile” problem in optical access network. Compared with wireless access network, it can solve the problem of shortage of communication range and bandwidth in wireless access network. This study is to discuss bandwidth allocation problem in optical network of hybrid access network. Because optical network in the bandwidth allocation will be the impact of wireless network, propose two different dynamic bandwidth allocation algorithms to reduce the impact. The first algorithm aims to improve the system utilization. The second algorithm aims to improve quality of low priority traffic on the premise that high priority traffic quality-of-service can be maintained.

碩士 / 國立交通大學 / 電控工程研究所 / 98 / For the surveillance, home security and medical applications a digital watermarking scheme is needed to provide full correlation which is trivial in the normal techniques when any attack is predicted to occur. The correlation of decoded data varies according to the variations of length of the embedding data for the type of attack predicted. We believe that proper selection of length for the embedding data for any watermarking system can assure full correlation. In this thesis, we propose a framework which adds a preprocessor and a splitter for any regular watermark scheme at the embedder and a combiner at the decoder. A length boundary database is constructed with the congregation of various lookup tables which contain the maximum length values for various types of attacks predicted. In the framework, the preprocessor obtains the maximum length for the type of attack predicted from the length boundary database and checks this length with that of the embedding data. If the embedding data length is more than the maximum length obtained, the splitter divides it into pieces of data each of length equal to that of the maximum length chosen. The embedding data is thus embedded with the acceptable length to obtain full correlation.

碩士 / 國立交通大學 / 機械工程學系 / 98 / Fall related injuries are serious public health issues all over the world, including in many developed countries, and most fall incidents occur at home. However, floor transition is a common yet frequently neglected condition when consider risk factors of falls. In consideration of high incident rates of fall-at-home events, our study aims to examine the effect of floor material change (transition), e.g. from living room to kitchen, on the control of level walking and gait stability, in order to reduce the floor material-induced risks of falls and improve the safety of home environment. We used three commonly-used floor materials at home in Taiwan – plane-wood, rough-ceramic, and polyvinyl chloride (PVC), to construct seven different floor conditions for the experiment. An eight-camera motion capture system and two force platforms were used to obtain kinematics and kinetics during level walking. Five healthy young adults and two older subjects volunteered for this study. All subjects completed 20 walking trials, on a 6-m long walkway at self-selected comfortable speed, each under two consistent condition (i.e. same floor material) – plane-wood and rough-wood floor, and four transition condition (i.e. from one material to another) – plane-wood to PVC, PVC to plane-wood, rough-wood to PVC and PVC to rough-wood floor. The five young participants also completed 20 trials on an additional dry to wet ceramic condition. Our results showed that when subjects walked from one floor to another with slightly lower coefficient of friction (COF), the peak required COF under foot increased 5% to 10%, and the foot angle at heel strike decreased about 30%. Moreover, peak anterior pelvic tilt angle was smaller as compared to those when walking under the consistent condition. The results imply a higher risk of falls during walking in transition than consistent conditions. Furthermore, the larger medial-lateral COM excursion also indicates that subjects had less stable locomotion under transition condition. Therefore, we suggest people avoid or use caution when they walk under transition condition, especially for the living environment of the elderly.

博士 / 國立交通大學 / 機械工程學系 / 98 / Zinc oxide (ZnO) thin film was deposited on glass substrate by RF reactive magnetron sputtering. The plasma parameters, structural, chemical, optical and hydrophilic/hydrophobic properties of the film were measured using a Langmuir probe, x-ray diffractometry (XRD), X-ray Photoelectron Spectroscopy (XPS), a UV-VIS spectrophotometer, and contact angle measurement, respectively. Results show that plasma density, electron temperature, deposition rate, and estimated ion bombardment energy all increased with increasing applied RF power. The deposition rate and ion bombardment energy were both influenced during ZnO thin film growth. Based on the measurements above, there are three distinct power regimes, which are strongly correlated with plasma properties. In the low power regime (50-100 W), a slow deposition rate produced the largest grain size. The smallest grain size appeared in the medium power regime (100-200 W). This is attributed to insufficient time for the adatoms to migrate on the substrate surface. In the high power regime (200-400 W), a relatively larger grain size appeared due to very large ion bombardment energy, which heated up the substrate and enhanced the thermal migration of adatoms. The optimal level of RF power for obtaining the optimum material properties of ZnO thin film is RF 400 W. However, the average transmittance is over 85% in the visible region of 400-700 nm for both pure ZnO thin film and ZnO on glass. This thesis also shows that ZnO thin film possessed better UV-shielding characteristics. For all of these optical properties, the thickness effect was more important than the power effect. The ZnO thin film exhibited hydrophobic characteristic when coated on glass substrate. For the O2/(Ar+O2) ratio effect, the results of this study show that the deposition rate decreased with an increasing O2/(Ar+O2) ratio. This was caused by a reduction in plasma density due to the addition of electronegative oxygen. At an O2/(Ar+O2) ratio of ~0.3, XRD results show that ZnO thin film forms a polycrystalline structure with the lowest FWHM (largest grain size). At the same time, the content fraction of the OⅠ peak (O-Zn bond) and the corresponding binding energy reached their highest levels at this O2/(Ar+O2) ratio, according to XPS analysis. The smallest FWHM, largest grain size, and less relaxation stress were simultaneously created by annealing treatment. The moderation annealing temperature is an important factor for improving material properties such as film quality, stress, grain size, transmittance, and UV-shielding characteristics, etc. An annealing treatment below 300℃ offers significant improvements, whereas it has no obvious variance when the annealing temperature exceeds 400℃. Finally, we deduced that the best annealing temperature is at 400℃ for glass substrate. In summary, we have obtained the excellent properties of ZnO thin film on glass with the high transparency, good UV-shielding, and hydrophobic characteristics for glass industry applications. The final section of this thesis provides directions for future research on different deposition parameters (distance between target and substrate, working pressure effect, etc.).

碩士 / 國立交通大學 / 電機學院碩士在職專班電機與控制組 / 98 / Currently, Field Sequential Color LCD (FSC-LCD) is the most eye-catching display technology. As the FSC-LCD does not require color filter (CF), it significantly reduces the cost of panels, and has the advantages of small size, high resolution, high color saturation, low power consumption and high luminous efficiency. However, the users do not accept this product owing to the color breakup (CBU) phenomenon which affects the quality of FSC image. The proposed “Active - liquid crystal and backlight (LC/BL) Control” algorithm presents the dynamic field sequential color (D-FSC) algorithm to reduce the CBU effect without greatly increasing frame rate. The D-FSC algorithm can intelligently select one adequate color sequence from multiple color sequences according to the image data. In other words, the scope of CBU suppression of the proposed Active -LC/BL Control algorithm is more extensive than other conventional FSCs. Besides, high-contrast and low-power image can be achieved by the two-dimension (2-D) LED dimming technique. But, the requirement of hardware is expensive. On the contrary, one pseudo zero-dimension (0-D) LED dimming technique can further reduce power dissipation with the D-FSC technique. Experimental results show the CBU suppression can be improved substantially by the proposed evaluation equation and the power dissipation can be reduced at least 12.5%.

碩士 / 國立交通大學 / 光電工程學系 / 98 / In this thesis we report two new application cases for the Artificial Neuron Network they are focus on the laser plus characterization and the locating of light spot. By using the learning ability and the function approximation, we report a method to retrace the phase in the frequency domain of the origin light field by the second harmonic generate spectrum, and also provide a possible experimental setup for realizing this measurement. On the other hand, by training Artificial Neuron Network to learn the position of a light spot within an image and combining the search method, the location of multiple light spot can be realized.

碩士 / 國立交通大學 / 電機學院通訊與網路科技產業專班 / 98 / This thesis introduces some topics about initial downlink synchronization, algorithm derivations, and program simulations of IEEE 802.16m system. When a mobile station entering to the network, it needs to perform initial synchronization, including of symbol timing offset, carrier frequency offset and preamble index. We utilize the trait which the power of preamble is larger than it of the common data symbol to compute the moving power sum, and then estimate the left boundary of preamble by finding out the peak value of moving power sum. A symbol period from this estimated boundary is regarded as the estimated preamble, which has a phase noise with the exact preamble. We derive the quasi maximum likelihood estimation from the likelihood function of the estimated preamble to obtain fractional carrier frequency offset (FCFO) and the formula of channel estimation. After compensating the estimated fractional carrier frequency offset to the formula of channel estimation, we substitute several reasonable integral carrier frequency offsets (ICFOs) and primary advanced preambles (PA-Preambles) into this formula and obtain channel impulse responses (CIRs). After that, we compute different fine timing offset index 64-points power sum of these CIRs and find out the peak value whose ICFO, PA-Preamble index, and fine timing offset index are regarded as the result of the joint estimation. In terms of simulations, above all, we present a reasonable received signal model of IEEE 802.16m, and then simulate our algorithm in floating point under AWGN (additive white Gaussian noise) channel to verify its performance. Moreover, we test this algorithm under different multipath fading channels, Pedestrian B and SUI-5, at different speed which cause Doppler shifts, and signal to noise ratio (SNR) to observe the effect on this algorithm.

碩士 / 國立交通大學 / 生醫工程研究所 / 98 / Researchers in the area of neural science, psychophysiology, communication and management technology believe that emotion plays an important role in decision making, learning and memory. In recent years, due to the development of instrument and experimental paradigm, emotion and its’ influence on decision making has become clear. This paper conducted an experiment based on Iowa Gambling Task (IGT) and utilized Artificial Neural Network (ANN) on Heart Rate Variability (HRV) analysis, and found out from between subject analysis that most subjects tend to make advantageous decisions rather than disadvantageous decisions when their heart rate slows down (p < 0.001). From within subjects analysis, through the difference of Inter Beat Interval (IBI), guessing accuracy of artificial neural network on the outcome of subjects’ decisions has reached above 65% (higher than 50%, which stands for guessing at random, p < 0.05), which means that there might exist relationships between the performance of human decision making and the psychophysiological activities appeared right before they made the decisions. The results might leads to a new path of development of decision support techniques based on the understanding of human emotions’ effects.