Corpus-Based Speech Enhancement With Uncertainty Modeling and Cepstral Smoothing

We present a new approach for corpus-based speech enhancement that significantly improves over a method published by Xiao and Nickel in 2010. Corpus-based enhancement systems do not merely filter an incoming noisy signal, but resynthesize its speech content via an inventory of pre-recorded clean signals. The goal of the procedure is to perceptually improve the sound of speech signals in background noise. The proposed new method modifies Xiao's method in four significant ways. Firstly, it employs a Gaussian mixture model (GMM) instead of a vector quantizer in the phoneme recognition front-end. Secondly, the state decoding of the recognition stage is supported with an uncertainty modeling technique. With the GMM and the uncertainty modeling it is possible to eliminate the need for noise dependent system training. Thirdly, the post-processing of the original method via sinusoidal modeling is replaced with a powerful cepstral smoothing operation. And lastly, due to the improvements of these modifications, it is possible to extend the operational bandwidth of the procedure from 4 kHz to 8 kHz. The performance of the proposed method was evaluated across different noise types and different signal-to-noise ratios. The new method was able to significantly outperform traditional methods, including the one by Xiao and Nickel, in terms of PESQ scores and other objective quality measures. Results of subjective CMOS tests over a smaller set of test samples support our claims.

Detecting Improvised Explosive Devices Via Forward Looking Ground Penetrating Radar

Forward-looking ground penetrating radar shows promise for detection of improvised explosive devices in active war zones. Because of certain insurmountable physical limitations, post-processing algorithm development is the most popular research topic in this field. One such investigative avenue explores the worthiness of frequency analysis during data post-processing. Using the finite difference time domain numerical method, simulations are run to test both mine and clutter frequency response. Mines are found to respond strongest at low frequencies and cause periodic changes in ground penetrating radar frequency results. These results are called into question, however, when clutter, a phenomenon generally known to be random, is also found to cause periodic frequency effects. Possible causes, including simulation inaccuracy, are considered. Although the clutter models used are found to be inadequately random, specular reflections of differing periodicity are found to return from both the mine and the ground. The presence of these specular reflections offers a potential alternative method of determining a mine’s presence.

Effect Of Confined Impinging Jet Mixing Processing Parameters On Poly (Lactic Acid) Nanoparticle Morphology

Biodegradable nanoparticles are at the forefront of drug delivery research as they provide numerous advantages over traditional drug delivery methods. An important factor affecting the ability of nanoparticles to circulate within the blood stream and interact with cells is their morphology. In this study a novel processing method, confined impinging jet mixing, was used to form poly (lactic acid) nanoparticles through a solvent-diffusion process with Pluronic F-127 being used as a stabilizing agent. This study focused on the effects of Reynolds number (flow rate), surfactant presence in mixing, and polymer concentration on the morphology of poly (lactic acid) nanoparticles. In addition to looking at the parameters affecting poly (lactic acid) morphology, this study attempted to improve nanoparticle isolation and purification methods to increase nanoparticle yield and ensure specific morphologies were not being excluded during isolation and purification. The isolation and purification methods used in this study were centrifugation and a stir cell. This study successfully produced particles having pyramidal and cubic morphologies. Despite successful production of these morphologies the yield of non-spherical particles was very low, additionally great variability existed between redundant trails. Surfactant was determined to be very important for the stabilization of nanoparticles in solution but appears to be unnecessary for the formation of nanoparticles. Isolation and purification methods that produce a high yield of surfactant free particles have still not been perfected and additional testing will be necessary for improvement.¿

Maximizing Product Formation and Minimizing Degradation: A Look at Buffering Conditions and Post-Reaction Degradation for the In-Line Jaffe Reaction with Capillary Electrophoresis

Creatinine levels in blood serum are typically used to assess renal function. Clinical determination of creatinine is often based on the Jaffe reaction, in which creatinine in the serum reacts with sodium picrate, resulting in a spectrophotometrically quantifiable product. Previous work from our lab has introduced an electrophoretically mediated initiation of this reaction, in which nanoliter plugs of individual reagent solutions can be added to the capillary and then mixed and reacted. Following electrophoretic separation of the product from excess reactant(s), the product can be directly determined on column. This work aims to gain a detailed understanding of the in-capillary reagent mixing dynamics, in-line reaction yield, and product degradation during electrophoresis, with an overall goal of improving assay sensitivity. One set of experiments focuses on maximizing product formation through manipulation of various conditions such as pH, voltage applied, and timing of the applied voltage, in addition to manipulations in the identity, concentration, and pH of the background electrolyte. Through this work, it was determined that dramatic changes in local voltage fields within the various reagent zones lead to ineffective reagent overlapping. Use of the software simulation program Simul 5 enabled visualization of the reaction dynamics within the capillary, specifically the wide variance between the electric field intensities within the creatinine and picrate zones. Because of this simulation work, the experimental method was modified to increase the ionic strength of the creatinine reagent zone to lower the local voltage field, thus producing more predictable and effective overlap conditions for the reagents and allowing the formation of more Jaffe product. As second set of experiments focuses on controlling the post-reaction product degradation. In that vein, we have systematically explored the importance of the identity, concentration, and pH of the background electrolyte on the post-reaction degradation rate of the product. Although prior work with borate background electrolytes indicated that product degradation was probably a function of the ionic strength of the background electrolyte, this work with a glycine background electrolyte demonstrates that degradation is in fact not a function of ionic strength of the background electrolyte. As the concentration and pH of the glycine background increased, the rate of degradation of product did not change dramatically, whereas in borate-buffered systems, the rate of Jaffe product degradation increased linearly with background electrolyte concentration above 100.0 mM borate. Similarly, increasing pH of the glycine background electrolyte did not result in a corresponding increase in product degradation, as it had with the borate background electrolyte. Other general trends that were observed include: increasing background electrolyte concentration increases peak efficiency and higher pH favors product formation; thus, it appears that use of a background electrolyte other than borate, such as glycine, the rate of degradation of the Jaffe product can be slowed, increasing the sensitivity of this in-line assay.