China 
Africa 
Explore chapters and articles related to this topic
Speech Recognition
Published in Sadaoki Furui, Digital Speech Processing, Synthesis, and Recognition, 2018
If a correlation structure between parameters can be established, and the correlation parameters can be estimated when training the general models, the parameters of unseen units can be adapted accordingly (Furui, 1980; Cox, 1995). To improve adaptation efficiency and effectiveness along this line, several techniques have been proposed, including probabilistic spectral mapping (Schwartz et al., 1987), cepstral normalization (Acero et al., 1990), and spectrum bias and shift transformation (Sankar and Lee, 1996). In addition to clustering and smoothing, a second type of constraint can be given to the model parameters so that all the parameters are adjusted simultaneously according to a predetermined set of transformations, e.g., a transformation based on multiple regression analysis (Furui, 1980). Various methods have recently been proposed in which a linear transformation (Affine transformation) between the reference and adaptive speaker-feature vectors is defined and then translated into a bias vector and a scaling matrix, which can be estimated using an EM algorithm (MLLR; Maximum Likelihood Linear Regression method) (Leggetter and Woodland, 1995). The transform parameters can be estimated from adaptation data that form pairs with the training data.
60 Molecular Junctions
Published in Tamar Seideman, Current-Driven Phenomena in Nanoelectronics, 2016
G. Schulze, K. J. Franke, J. I. Pascual
Figure 5.6 (a) STM image of an island of C60 molecules on Pb(111) before and after performing I(Z) events on 29 noncontiguous molecules (chosen to form a 2 x 2 lattice). After that procedure, all indented molecules show an apparent lower height and, in most cases, their characteristic resonant fingerprint disappears from their spectrum (bias during indentation Vs = 2.25 V). The lower image has been filtered using the freeware WSxM (Ref. 35) to improve the visualization of degraded fullerenes.
Experience of drivers of all age groups in accepting autonomous vehicle technology
Published in Journal of Intelligent Transportation Systems, 2023
Sherrilene Classen, Virginia P. Sisiopiku, Justin R. Mason, Wencui Yang, Seung-Woo Hwangbo, Brandy McKinney, Yuan Li
As such, this study has inherent biases, such as a self-selection bias, spectrum bias, Hawthorne bias (i.e., the presence of a safety operator, required by NHTSA, in the shuttle), and demand characteristics (i.e., the effect of word-of-mouth referral on enrollment). Even though we have tried to cast the recruitment net far and wide, we can at best describe the sample as a convenience sample. Therefore, this study’s findings, although they provide foundational knowledge for the AV technology industry, are only generalizable to study participants and settings that fit the demographic profile and context of this study.