Assessment of a patients risk of committing suicide is important in order for them to be able to receive early hospitalization. The two critical mental states that may arouse much confusion to the clinicians are the high risk suicidal patients and the depressed patients. Current diagnostic tools are time-consuming and require the clinicians to have proper training and experience in order to obtain useful assessment and sufficient information for determining suitable treatments. A lot of information regarding the psychological state is accessible through the human voice content. Our desire is to show the effectiveness of using the Power Spectrum Density of male speech as a potential feature for distinguishing between the high risk suicidal patients and the depressed patients. The classification analyses were performed using the linear and quadratic classifiers together with the three methods of resampling (Equal-Test-Train, Jackknife and Cross-validation). According to the classification results, after removing outlier patients, the feature was shown to be quite effective in identifying between the high risk suicidal patients and the depressed patients.

Suicide is a major public health problem in the US. The procedure to measure the degree of suicidal risk in depressed patients is complicated and time consuming. Therefore, there is a need to develop a diagnostic tool to aid physicians in determining suicidal risk. Speech has been identified to be able to reflect emotional conditions including depression and suicidal thoughts. This paper represents one of the development steps in creating a speech-based diagnostic tool to help physicians to make clinical judgments. It analyzes the acoustic features based on the power spectral density extracted from the speech of female patients in order to detect suicidal risk. The focus of the experiments is on the classification between depressed and high-risk female patients. Two types of speech, spontaneous and automatic, were analyzed independently using multiple statistical approaches and the classification results are discussed. Possible outliers are observed in the spontaneous speech analysis while the automatic speech analysis produces satisfactory classification results. It is shown that speech features can be used as an indicator for suicidal risk of female patients. Potential future work that would advance our knowledge are also proposed.

A typical image-guided neurosurgery system (IGNS) presents images acquired pre-operatively to the surgeons to assist in planning and executing the procedure. The major issue with IGNSs is the fact that the brain shifts during the procedure due to a number of reasons ranging from loss of cerebrospinal fluid, medications, and/or resection of abnormal tissues. Updating the pre-operative images to compensate for brain shift using computational models is an active area of research. In this dissertation, we have developed and evaluated techniques that permit estimating cortical displacements from laser range scanner (LRS) images and intra-operative microscope video sequences. Cortical displacements can be used subsequently as input to computational models. The set of techniques that were developed include novel vessel segmentation algorithms, new registration methods for pre- and post-resection LRS data and new tracking algorithms for microscope video sequences recorded during a tumor resection surgery. Results show that combining LRS images and video sequences is a promising approach to estimate intra-operative brain shift.

Power is a critical issue in digital system design, especially with the emphasis on the portability of electronic devices. However, decreasing power does not necessarily lead to energy efficiency; designers need to consider the negative influence on performance when power is reduced. Trade-offs in circuit design should be evaluated using both power and performance. One important element of power consumption in a digital system is the flip-flop. This thesis surveys several previous designs of double edge-triggered flip-flops, and then proposes a transmission-gate-based, double edge-triggered flip-flop with a novel clock gating function. Two designs (with and without clock gating) are each compared against two benchmark circuit designs. Using the second benchmark circuit (from the literature), the Design II: P_DETFF with the clock gating function saved 33% power on average when the switching activity factor (α) ranged from 0 to 0.4. When the input is idle, it also can save up to 98% of the power compared to the baseline Benchmark I: SETFF. The Design I: T_DETFF showed better performance than any other designs when α was above 0.4. Comparing with the best benchmark in this range, it saved 23% of the power on average, and 27% when the input switches every clock cycle. The proposed designs required slightly more area than the benchmarks, but maintained performance across different frequencies.

High confidence embedded control system software often requires formal analyses to ensure design correctness. Detailed models cover numerous design concerns such as controller stability, timing requirements, fault tolerance, and deadlock freedom. Models for each of these design domains must together provide a consistent and faithful representation of the potential problems an operational system would face. Coupling between separately designed components and modules prevents model analyses from scaling well to large designs. Coupling also occurs within individual systems and components between behaviors represented by different design concerns as different aspects of the design constrain design structures and parameters in conflicting ways. These complications combine with other factors to increase the difficulty of system integration, leading to high costs and long development schedules. Correctness properties in model-based designs take one of the following forms (from Edwards et al): 1) Properties inherent to the model of computation, which hold for all valid specifications, 2) Syntactic properties which can be determined by analysis of the structure of specification elements, and 3) Semantic properties which require examination of the actual behavior of the specification. To tackle problems associated with highly coupled designs, their required correctness constraints, and the cost impact on design projects, we have created the Embedded Systems Modeling Language (ESMoL). Well-formed models in ESMoL inherently express functional determinism (from dataflow semantics), deadlock-freedom (from synchrony), and timing determinism (from a time-triggered platform). To address more complex syntactic and semantic properties we propose incremental analysis, which allows us to extend previously calculated design analysis results to new features in an efficient way. We give an example of incremental syntactic analysis in the form of cycle checking to ensure well-formedness of ESMoL models. We give an example of incremental semantic analysis in scheduling dataflow graphs subject to end-to-end latency constraints. Effective incremental analysis techniques can allow developers to more rapidly iterate and converge on a correct design.

Bias circuits (e.g. current sources) provide essential global signals in analog and mixed-signal design. Ideally, a bias circuit should be invariant over operating conditions such as temperature, output load, and supply voltage. Given the effort and cost required to implement a high-performance precision bias current source, current mirrors are typically used to replicate a single stable current throughout an entire integrated circuit. Consequently, a single-event (SE) strike to a critical bias circuit node can have wide-ranging global effects throughout the IC. A hardened precision bias current source is essential to prevent multiple errors from disrupting the operation of an entire integrated system. This thesis presents a novel radiation-hardened-by-design (RHBD) technique that takes advantage of the multi-node charge collection mechanism and employs it through a balancing and mirroring circuit topology to mitigate the effects of a single event strike. This technique, called sensitive node active charge cancellation (SNACC), can be applied to harden critical nodes in analog and mixed-signal circuits. In this work, the SNACC technique is applied to a bias current source topology typical of the designs used throughout industry. The hardened bias circuit is compared with a traditional capacitive hardening technique to quantify its usefulness and performance. The SNACC hardening technique is verified using simulations in a 90-nm CMOS process.

Soft errors are becoming an increasingly important issue in todays microelectronics industry. With decreasing transistor sizes and increasing transistor counts, it is simply not efficient to harden every circuit module against soft errors. Thus a designer must selectively harden the most vulnerable modules of a design. This thesis addresses the need for a non-invasive, portable, and easy to use software tool for determining the architectural-vulnerability factor (AVF) of a given design and its sub-modules. Fault injection and analysis is carried out at the register-transfer level to cut down on simulation time and simulation space. The Verilog-procedural interface (VPI) is used to implement the fault injection and error detection. Testing has been carried out on an ASIC design with 1.1 million flip-flops and an eight-bit microprocessor. Simulation results for 1,500 runs of the ASIC and 50,000 runs of the microprocessor are used to estimate AVF for the designs and their sub-modules. Other relevant results to help determine the vulnerability of a circuit module, such as error latency, benign errors, and silent errors, are also evaluated. A designer can then use this data to select the most vulnerable sub-modules or architectural structures of a design for hardening against soft errors.

Single-event upsets and errors are of growing concern as technology scales toward smaller transistor sizes. While smaller transistors allow for greater on-chip integration, this comes with the penalties of reduced supply voltage overhead and low drive currents compared to larger technologies. These penalties provide added challenges when considering the use of state of the art technologies for space based and strategic analog / mixed-signal applications. Therefore, it may prove beneficial to consider the use of slightly older technologies that avoid these penalties for such applications.<p>In this study, the general usability of the 180 nm technology in a space environment setting was explored through simulation with an emphasis on analog / mixed-signal applications. While a bit dated, limited published data exists on this technologys response to single events. Therefore, simulations were performed across variations in supply voltage, LET, and transistor load to generally characterize the technologys susceptibility to single-event transients and single-event latchup. General observed trends are reported for these phenomena along with the effects of commonly used mitigation techniques including highly doped buried layers, guard rings, and triple-well NMOS devices.

The human active visual perception is closely related to a fixation-move-fixation rhythm. In recent past, the scientists adopted this concept with the expectation that camera-equipped robot can interact with the environment better by altering the viewpoint of cameras rather than passively observing the environment. In this thesis, a pair of pan-tilt cameras mounted on the ISAC humanoid robot was used to implement an active vision process along a one-dimensional search space. The software system including an object recognition sub-system was developed and was integrated with an arm control sub-system. The object recognition sub-system was designed based on color and shape detection and was implemented by using OpenCV libraries and MATLAB. Since two cameras were moving during the experiments, there was a linear mapping underlying the different combinations of angles of two sets of pan-tilt units for object localization. Several experiments were performed to evaluate system performance.

Since their discovery, phosphors have been studied extensively and have pervaded our everyday lives. Despite the wealth of research conducted on phosphor powders there has been no research published on long-term effects of radiation-induced damage within the photoluminescence spectra. Phosphors have the potential for the use as radiation sensors, where they could conceivably be distributed over a surface and queried from afar. This work establishes the sensitivity of the luminescence spectra of two different phosphor materials to different types and amounts of exposure. Changes in photoluminescence spectra were quantified by comparing pre and post exposure intensity measurements and emission curves for each material, synthesis method and exposure type. Trends for the changes in photoluminescence intensity are reported along with data from variability measurements.