In this paper, we present a comparative analysis of artificial neural networks (ANNs) and Gaussian mixture models (GMMs) for design of voice conversion system using line spectral frequencies (LSFs) as feature vectors. Both the ANN and GMM based models are explored to capture nonlinear mapping functions for modifying the vocal tract characteristics of a source speaker according to a desired target speaker. The LSFs are used to represent the vocal tract transfer function of a particular speaker. Mapping of the intonation patterns (pitch contour) is carried out using a codebook based model at segmental level. The energy profile of the signal is modified using a fixed scaling factor defined between the source and target speakers at the segmental level. Two different methods for residual modification such as residual copying and residual selection methods are used to generate the target residual signal. The performance of ANN and GMM based voice conversion (VC) system are conducted using subjective and objective measures. The results indicate that the proposed ANN-based model using LSFs feature set may be used as an alternative to state-of-the-art GMM-based models used to design a voice conversion system. 相似文献
A tricyclazole selective chitosan/Fe3O4 magnetic molecularly imprinted polymer (MMIP) was synthesized using non-covalent binding polymerization involving methacrylic acid (MAA) as functional monomer, divinylbenzene (DVB-80) as crosslinker, 2,2'-azobisisobutyronitrile as initiator, acetonitrile/toluene (75:25, v/v) as porogenic solvent and tricyclazole as template. Surface morphology and magnetic characterization of the prepared imprinted and non-imprinted polymers were done using scanning electron microscopy, transmission electron microscopy, Fourier transform infrared spectrometry and vibrating sample magnetometry, respectively. The adsorption kinetic data fitted best in pseudo-second-order model. The adsorption equilibrium was achieved in 30 min and the maximum binding capacity was 4579.9 µg/g. The Freundlich isotherm model was found suitable for explaining the binding isotherm data (R2 > 0.99). Negative values of thermodynamic parameters ∆G (Gibb’s free energy), ∆H (enthalpy), and ∆S (entropy) revealed exothermic and spontaneous nature of adsorption processes. It also revealed decreased randomness at the solid–liquid interface during sorption. The scatchard plot analysis suggested heterogeneity of binding sites on MMIPs. The molecular recognition selectivity of MMIPs towards tricyclazole was much higher, as compared to its structural analogues, tebuconazole (α = 28.58) and hexaconazole (α = 37.16). The MMIPs were successfully applied to separate and enrich tricyclazole from fortified samples of rice and water, with a recovery percentage of 89.4% and 90.9%, respectively. These reusable imprinted polymers possessing high selectivity and specificity can be utilized as an adsorbent for solid-phase extraction in sample preparation for tricyclazole residue analysis in complex environmental matrices.
The short backfire antenna (SBA) has been widely used for mobile satellite communications, tracking, telemetry, and wireless local-area network applications due to its compact structure and excellent radiation characteristics. The most common excitation topology for the SBA is a balance-fed wire dipole, which has the disadvantage of a narrow frequency bandwidth for the input impedance. In this paper, an H-shaped slot is employed to excite the SBA for the first time. The H-shaped slot is unbalance-fed from a coaxial line. It is demonstrated that the H-shaped slot-excited SBA can achieve a bandwidth for input impedance of more than 20% (VSWR<2) while maintaining good radiation performance. The antenna structure is described and the simulation and experimental results are presented. The operating principle is investigated to explain why the slot-excited SBA can result in good impedance and radiation characteristics. A parametric study is conducted for the use of practical engineering design. 相似文献
This paper presents a novel experimental approach for the analysis of factors to be considered when designing a flip-chip package. It includes the design of an experiment and statistical analysis of the outputs and uses both test-structure measurements and full-wave simulation techniques in the 1-35-GHz frequency range. The most significant factors are found to be, from the most to least important, the length of the area where the device and substrate overlap (referred to as conductor overlap), the bump diameter, and the width of the coplanar-waveguide transmission-line launch. These results are valid for conductor overlaps between 300-500 μm. For a lower value (120 μm), the significance level of the overlap decreases and the bump height also becomes significant. Test-structure measurements in the 120-200-μm overlap range validate this result and demonstrate the decrease in the significance level. The substrate thickness in the 10-25-mil interval is found to be statistically insignificant, therefore, it can be eliminated from further analysis. This approach provides a foundation for development of a set of design rules for RF and microwave flip-chip similar to RF integrated-circuit design rules 相似文献
A micro-power complementary metal oxide semiconductor (CMOS) low-noise amplifier (LNA) is presented based on subthreshold MOS operation in the GHz range. The LNA is fabricated in an 0.18-/spl mu/m CMOS process and has a gain of 13.6 dB at 1 GHz while drawing 260 /spl mu/A from a 1-V supply. An unrestrained bias technique, that automatically increases bias currents at high input power levels, is used to raise the input P1dB to -0.2 dBm. The LNA has a measured noise figure of 4.6 dB and an IIP3 of 7.2 dBm. 相似文献
In this paper, three-dimensional (3-D) integrated cavity resonators and filters consisting of via walls are demonstrated as a system-on-package compact solution for RF front-end modules at 60 GHz using low-temperature cofired ceramic (LTCC) technology. Slot excitation with a /spl lambda/g/4 open stub has been applied and evaluated in terms of experimental performance and fabrication accuracy and simplicity. The strongly coupled cavity resonator provides an insertion loss <0.84 dB, a return loss >20.6 dB over the passband (/spl sim/0.89 GHz), and a 3-dB bandwidth of approximately 1.5% (/spl sim/0.89 GHz), as well as a simple fabrication of the feeding structure (since it does not require to drill vias to implement the feeding structure). The design has been utilized to develop a 3-D low-loss three-pole bandpass filter for 60-GHz wireless local area network narrow-band (/spl sim/1 GHz) applications. This is the first demonstration entirely authenticated by measurement data for 60-GHz 3-D LTCC cavity filters. This filter exhibits an insertion loss of 2.14 dB at the center frequency of 58.7 GHz, a rejection >16.4 dB over the passband, and a 3-dB bandwidth approximately 1.38% (/spl sim/0.9 GHz). 相似文献
The observation of negative differential resistance (NDR) and negative transconductance at high drain and gate fields in depletion-mode AlGaAs/InGaAs/GaAs MODFETs with gate lengths Lg ~0.25 μm is discussed. It is shown that under high bias voltage conditions, Vds>2.5 V and Vgs>0 V, the device drain current characteristic switches from a high current state to a low current state, resulting in reflection gain in the drain circuit of the MODFET. The decrease in the drain current of the device corresponds to a sudden increase in the gate current. It is shown that the device can be operated in two regions: (1) standard MODFET operation for Vgs<0 V resulting in fmax values of >120 GHz, and (2) a NDR region which yields operation as a reflection gain amplifier for Vgs >0 V and Vds>2.5 V, resulting in 2 dB of reflection gain at 26.5 GHz. The NDR is attributed to the redistribution of charge and voltage in the channel caused by electrons crossing the heterobarrier under high-field conditions. The NDR gain regime, which is controllable by gate and drain voltages, is a new operating mode for MODFETs under high bias conditions 相似文献
A 60-GHz fully integrated bits-in bits-out on–off keying (OOK) digital radio has been designed in a standard 90-nm CMOS process technology. The transmitter provides 2 dBm of output power at a 3.5-Gb/s data rate while consuming 156 mW of dc power, including the on-chip 60-GHz frequency synthesizer. A pulse-shaping filter has been integrated to support high data rates while maintaining spectral efficiency. The receiver performs direct-conversion noncoherent demodulation at data rates up to 3.5 Gb/s while consuming 108 mW of dc power, for a total average transceiver energy consumption of 38 pJ/bit in time division duplex operation. To the best of the authors' knowledge, this is the lowest energy per bit reported to date in the 60-GHz band for fully integrated single-chip CMOS OOK radios. 相似文献
Multimedia Tools and Applications - In the recent digitization era, image hashing is a key technology, including image recognition, authentication and manipulation detection, among many multimedia... 相似文献
The redundant data in multichannel electroencephalogram (EEG) signals significantly reduces the performance of brain–computer interface (BCI) systems. By removing redundant channels, a channel selection strategy increases the classification accuracy of BCI systems. In this work, a novel channel selection method (stdWC) based on the standard deviation of wavelet coefficients across channels is proposed to identify Motor Imagery (MI) based EEG signals. The wavelet coefficients are calculated by employing a Continuous Wavelet Transform (CWT) filter bank to decompose each trial from the EEG channel. The wavelet coefficient's standard deviation values are obtained across the channels, and these values are then sorted to determine the EEG channels with the highest standard deviation values. The channels with the largest wavelet coefficient divergence are chosen. MI trials are then spatially filtered with the Common Spatial Pattern (CSP), and CWT filter bank-based 2D images are generated from the spatially filtered trials. These images are then classified using a unique nine-layered convolutional neural network (CNN) model that combines two feature maps acquired with differing filter sizes. The proposed framework (stdWC-CSP-CNN) is evaluated using kappa score and classification accuracy on two publically accessible datasets (BCI Competition III dataset IVa and BCI Competition IV dataset 2a). The suggested framework achieved a mean test classification accuracy of 88.8% for dataset IVa from BCI Competition III and 75.03% for dataset 2a from BCI Competition IV, according to the results. The proposed channel selection method outperforms the other channel selection methods examined, according to the results. By rejecting redundant channels, the whole framework can improve the performance of MI-based BCIs. 相似文献