Analysis of Speaker Adaptation Algorithms for HMM-Based Speech Synthesis and a Constrained SMAPLR Adaptation Algorithm

In this paper, we analyze the effects of several factors and configuration choices encountered during training and model construction when we want to obtain better and more stable adaptation in HMM-based speech synthesis. We then propose a new adaptation algorithm called constrained structural maximum a posteriori linear regression (CSMAPLR) whose derivation is based on the knowledge obtained in this analysis and on the results of comparing several conventional adaptation algorithms. Here, we investigate six major aspects of the speaker adaptation: initial models; the amount of the training data for the initial models; the transform functions, estimation criteria, and sensitivity of several linear regression adaptation algorithms; and combination algorithms. Analyzing the effect of the initial model, we compare speaker-dependent models, gender-independent models, and the simultaneous use of the gender-dependent models to single use of the gender-dependent models. Analyzing the effect of the transform functions, we compare the transform function for only mean vectors with that for mean vectors and covariance matrices. Analyzing the effect of the estimation criteria, we compare the ML criterion with a robust estimation criterion called structural MAP. We evaluate the sensitivity of several thresholds for the piecewise linear regression algorithms and take up methods combining MAP adaptation with the linear regression algorithms. We incorporate these adaptation algorithms into our speech synthesis system and present several subjective and objective evaluation results showing the utility and effectiveness of these algorithms in speaker adaptation for HMM-based speech synthesis.     

Sensory evaluation of beef patties and sausages containing lipids with various component fatty acids

Sensory scores of beef patties and emulsion-type sausages containing varying levels of lipids with various component fatty acids were investigated. Sensory scores of beef patties containing solid fat were relatively constant at every fat level examined. However, scores of beef patties containing liquid oil decreased with increasing level of added oil. In the case of sausages, on the other hand, sensory evaluation scores markedly decreased with increasing the level of solid fat, while no significant change was observed in panel scores when liquid oil was mixed into sausages. Even 10% level of added lipids satisfactorily gave preferable acceptability to beef patties and sausages.     

Experimental and clinical study on intraoperative local infusion chemotherapy (ILIC) for advanced gastric carcinoma

An intensive chemotherapy combined with surgery, termed "intraoperative local infusion chemotherapy (ILIC)", was devised. The ILIC procedure is to infuse 50 mg of cisplatin through the feeding artery to the tumor, which has been isolated from blood flow by clamping the stomach and blood vessels, before radical surgery. An experiment in dogs showed considerable free Pt transfer into thoracic lymph after ILIC. Sum total Pt in the thoracic lymph after ILIC (294.5 micrograms/3 h) was 4.3 times that after IV administration (68.3 micrograms/3 h). In ILIC, AUC of free Pt in the thoracic lymph was about 5.8 times that in peripheral blood. ILIC was applied in 24 advanced gastric cancer patients, consisting of 6, 2, 4 and 12 in pTNM stage II, IIIA, IIIB and IV, respectively. Pt concentration of the tissues was high in the tumor (23.22 micrograms/g) and the regional lymph nodes (2.95 micrograms/g) compared to that in the serum (0.45 microgram/ml). The survival rate (Kaplan-Meier) in the ILIC patients was significantly higher than that of control patients (matched pair method) from among patients treated by surgery alone. No fatal complications were encountered in any patient treated by ILIC.     

Reduced brain serotonin transporter availability in major depression as measured by [123I]-2 beta-carbomethoxy-3 beta-(4-iodophenyl)tropane and single photon emission computed tomography

BACKGROUND: Prior research has suggested reductions in the density of serotonin transporter (SERT) binding sites in blood platelets and post-mortem brain tissue of depressed patients. We sought to determine whether patients with unipolar major depression have diminished SERT availability as assessed by both brainstem [123I] beta-CIT SPECT and platelet [3H]paroxetine binding. METHODS: Drug-free depressed and healthy subjects were injected with 211 +/- 22 MBq [123I] beta-CIT and imaged 24 +/- 2 h later under equilibrium conditions. A ratio of specific to nonspecific brain uptake (V3" = (brainstem-occipital)/occipital), a measure proportional to the binding potential (Bmax/Kd), was used for all comparisons. RESULTS: Results showed a statistically significant reduction in brainstem V3" values in depressed as compared to healthy subjects (3.1 +/- .9 vs. 3.8 +/- .8, p = .02). Platelet [3H]paroxetine binding was not altered (Bmax = 2389 +/- 484 vs. 2415 +/- 538 fmol/mg protein, p = .91) and was not significantly correlated with brainstem [123I] beta-CIT binding (r = -0.14, p = .48). CONCLUSIONS: These data are the first to suggest reductions in the density of brain SERT binding sites in living depressed patients. These findings provide further support for a preeminent role for alterations in serotonergic neurons in the pathophysiology of depression.     

A 63-year-old man with progressive cauda equina/conus medullaris syndrome

BACKGROUND: This article presents results of the acute treatment phase of a 2-site study comparing cognitive behavioral group therapy (CBGT) and treatment with the monoamine oxidase inhibitor phenelzine sulfate for social phobia. METHODS: One hundred thirty-three patients from 2 sites received 12 weeks of CBGT, phenelzine therapy, pill placebo administration, or educational-supportive group therapy (an attention-placebo treatment of equal credibility to CBGT). The "allegiance effect," ie, the tendency for treatments to seem most efficacious in settings of similar theoretical orientation and less efficacious in theoretically divergent settings, was also examined by comparing responses to the treatment conditions at both sites: 1 known for pharmacological treatment of anxiety disorders and the other for cognitive behavioral treatment. RESULTs: After 12 weeks, phenelzine therapy and CBGT led to superior response rates and greater change on dimensional measures than did either control condition. However, response to phenelzine therapy was more evident after 6 weeks, and phenelzine therapy was also superior to CBGT after 12 weeks on some measures. There were few differences between sites, suggesting that these treatments can be efficacious at facilities with differing theoretical allegiances. CONCLUSIONS: After 12 weeks, both phenelzine therapy and CBGT were associated with marked positive response. Although phenelzine therapy was superior to CBGT on some measures, both were more efficacious than the control conditions. More extended cognitive behavioral treatment and the combination of modalities may enhance treatment effect.     

The Internet & Healthcare Education: HELIX

With the advent of the World Wide Web (WWW), we are now on the cusp of a revolution in computer technology that will dramatically enhance medical education. An historical analogy might be Johann Gutenberg's invention of movable type in the 1400's-radically decreasing the cost, time, and expertise required to reproduce printed materials. Now, the WWW can decrease the cost of disseminating medical educational materials. When an educational module is authored for the Web, it can be placed on a computer "server" which in turn, distributes the program on the WWW to anyone with a computer and Internet access. Rapidly emerging standards are being developed to allow increasingly rich educational experiences on the Internet. With the introduction of HTML (hypertext markup language), a standardized method of placing text and graphics, as well as the connections between them, was created.     

An SNS-based model for finding collaborative partners

This paper proposes a model, Recommendation of Appropriate Partners (RAP), used on a Social Networking Service (SNS) for locating appropriate “helpers” for users based on individual users’ Chain of Friends (CoF) relationships. Using the RAP model, individual users can participate in a collaborative online community in remote locations, whereby helpers are willing to help other users solve their tasks/problems, and it is intended that both the users and helpers gain knowledge from these interactive online sessions. An example of the RAP-based system was implemented to invite Program Committee members to an international conference. The system was evaluated and the experimental results show that our model is very effective for discovering collaboration partners and finding users with similar interests in order to create communities for providing future and longer-term helping exchange.

     

Efficient blind dereverberation and echo cancellation based on independent component analysis for actual acoustic signals

This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to cope with reverberations. To extract a clean signal from the reverberant observation, we model the separation process in the short-time Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. Therefore, the main issue in dereverberation is to reduce the high computational cost of ICA. In this letter, we reduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that our method improves the word correctness of automatic speech recognition by 10 to 20 points in a RT??= 670 ms reverberant environment.     

Studies of the pH of the membrane surface in a laboratory chlor-alkali cell

The pH on the surface of an ion-exchange membrane was measured with a microprobe in a laboratory scale chlor-alkali cell to study the leakage of OH^– through the membrane. The solution pH in the vicinity of the membrane facing the anode was found to be considerably higher than that in the bulk solution because of penetration and leakage of OH^–. The pH varied with the membrane type. This explains why the carboxylate membrane is not protonated and can be used as a separator in chlor-alkali cells.Deceased.