scholarly journals Time-Domain Speech Extraction with Spatial Information and Multi Speaker Conditioning Mechanism

Author(s):  
Jisi Zhang ◽  
Catalin Zorila ◽  
Rama Doddipatla ◽  
Jon Barker
2021 ◽  
Author(s):  
Shuang Sha ◽  
Tingting Lyu ◽  
Hao Zhang ◽  
T. Aaron Gulliver

Abstract Background: Space-time adaptive processing (STAP) has been widely used in the fields of communication, radar, and navigation anti-jamming. However, the traditional scalar array used by STAP has certain limitations,because it can only obtain spatial information. In order to further improve the performance of the space-time domain joint filtering technology, this paper replaces the traditional scalar array with an alternating polarization sensitive array (APSA). Compared with the dual polarization sensitive array (DPSA), it can not only obtain the polarization information of the signal, but also reduce the computational complexity of the algorithm. Methods: Using the polarization information of the signals, this paper realizes an alternate polarization sensitive array space-time-polarization adaptive processing algorithm (APSA-STPAP) based on the linear variance minimum criterion (LCMV). Different from the traditional LCMV criterion, this paper takes the space-time polarization joint steering vector of the desired signal and the interference signal as the constraint matrix, and uses the "1 condition" and "zero condition" as the constraint conditions to effectively suppress the interference signal and enhance the expectation signal. Results: The simulation results show that: (1) APSA-STPAP algorithm can achieve the same filtering effect as DPSA-STPAP algorithm. From the perspective of the spatial domain, time domain and polarization domain, it can form null in the direction of interference, effectively suppress the interference signal, and realize space-time-polarization adaptive processing. (2) Under the same conditions, APSA-STPAP algorithm can achieve the same filtering effect as DPSA-STPAP algorithm. there is a big difference between the two algorithms, APSA-STPAP algorithm can effectively reduce the amount of computation. Moreover, the dipole of alternating polarization sensitive array is halved, which reduces the coupling effect between electric dipoles, and is conducive to engineering implementation. (3) APSA-STPAP algorithm can maintain good anti-interference performance even when the electric dipole and anti-jamming degree of freedom are reduced by half, and its anti-jamming performance is similar to that of polarization sensitive array. The output SINR of DPSA-STPAP algorithm is about 3dB higher than that of APSA-STPAP algorithm. There is little difference between the anti- interference performance of APSA and DPSA.


Author(s):  
Nitam Sunuwar ◽  
Jung-Ryul Lee

Interest in ultrasonic guided wave based Structural Health Monitoring and a nondestructive evaluation system has grown in recent years, especially to monitor thin plate like structures. However, an effective signal processing and imaging algorithms are essential to achieve necessary performance. This paper describes wave rich laser ultrasonic wavenumber imaging method (UWI) method for damage visualization. Ultrasonic waves were generated by a scanning laser source and acquired using a capacitance air coupled transducer (ACT). However, the inherent existence of multiple Lamb wave modes in signal makes it harder for effective damage evaluation. This is further complicated if the reflections from the boundaries are present in the signal. The use of an ACT with an in-line programmable filter helps to isolate lower order Lamb wave modes (Ao and So), since the dispersive waves radiate at certain angle from the specimen governed by Snell’s law. By comparing the results from the ultrasonic wavefield image obtained using the ACT and a PZT contact sensor under the same experimental condition, mode isolation phenomena was verified. Such isolated wave mode was processed using a proposed wave rich UWI algorithm where a wave rich field was generated by superposing the wavefields. The mode filtered measurements were arranged in 3D space-time domain where each slice in time domain represents 2D wavefield image. A 2D Fast Fourier Transform (FFT) was applied to this spatial information in time domain which transformed it to a wavenumber domain. A wavenumber filter is then applied and inverse Fourier transformed to get back to the wavenumber filtered measurement. However, instead of applying filter to every 2D slice in time domain, certain frames were selected and merged to replicate wave propagation in total scan-area. This wave rich field not only saves time and space but also reduce computational complexity during post-processing. This method was tested successfully in an aluminum plate with milled area damage and a composite fiber-reinforced plastic (CFRP) wing skin with two impact damages.


Author(s):  
T. A. Welton

Various authors have emphasized the spatial information resident in an electron micrograph taken with adequately coherent radiation. In view of the completion of at least one such instrument, this opportunity is taken to summarize the state of the art of processing such micrographs. We use the usual symbols for the aberration coefficients, and supplement these with £ and 6 for the transverse coherence length and the fractional energy spread respectively. He also assume a weak, biologically interesting sample, with principal interest lying in the molecular skeleton remaining after obvious hydrogen loss and other radiation damage has occurred.


Author(s):  
Vijay Krishnamurthi ◽  
Brent Bailey ◽  
Frederick Lanni

Excitation field synthesis (EFS) refers to the use of an interference optical system in a direct-imaging microscope to improve 3D resolution by axially-selective excitation of fluorescence within a specimen. The excitation field can be thought of as a weighting factor for the point-spread function (PSF) of the microscope, so that the optical transfer function (OTF) gets expanded by convolution with the Fourier transform of the field intensity. The simplest EFS system is the standing-wave fluorescence microscope, in which an axially-periodic excitation field is set up through the specimen by interference of a pair of collimated, coherent, s-polarized beams that enter the specimen from opposite sides at matching angles. In this case, spatial information about the object is recovered in the central OTF passband, plus two symmetric, axially-shifted sidebands. Gaps between these bands represent "lost" information about the 3D structure of the object. Because the sideband shift is equal to the spatial frequency of the standing-wave (SW) field, more complete recovery of information is possible by superposition of fields having different periods. When all of the fields have an antinode at a common plane (set to be coincident with the in-focus plane), the "synthesized" field is peaked in a narrow infocus zone.


Author(s):  
John R. Porter

New ceramic fibers, currently in various stages of commercial development, have been consolidated in intermetallic matrices such as γ-TiAl and FeAl. Fiber types include SiC, TiB2 and polycrystalline and single crystal Al2O3. This work required the development of techniques to characterize the thermochemical stability of these fibers in different matrices.SEM/EDS elemental mapping was used for this work. To obtain qualitative compositional/spatial information, the best realistically achievable counting statistics were required. We established that 128 × 128 maps, acquired with a 20 KeV accelerating voltage, 3 sec. live time per pixel (total mapping time, 18 h) and with beam current adjusted to give 30% dead time, provided adequate image quality at a magnification of 800X. The maps were acquired, with backgrounds subtracted, using a Noran TN 5500 EDS system. The images and maps were transferred to a Macintosh and converted into TIFF files using either TIFF Maker, or TNtolMAGE, a Microsoft QuickBASIC program developed at the Science Center. From TIFF files, images and maps were opened in either NIH Image or Adobe Photoshop for processing and analysis and printed from Microsoft Powerpoint on a Kodak XL7700 dye transfer image printer.


Author(s):  
RAD Mackenzie ◽  
G D W Smith ◽  
A. Cerezo ◽  
J A Liddle ◽  
CRM Grovenor ◽  
...  

The position sensitive atom probe (POSAP), described briefly elsewhere in these proceedings, permits both chemical and spatial information in three dimensions to be recorded from a small volume of material. This technique is particularly applicable to situations where there are fine scale variations in composition present in the material under investigation. We report the application of the POSAP to the characterisation of semiconductor multiple quantum wells and metallic multilayers.The application of devices prepared from quantum well materials depends on the ability to accurately control both the quantum well composition and the quality of the interfaces between the well and barrier layers. A series of metal organic chemical vapour deposition (MOCVD) grown GaInAs-InP quantum wells were examined after being prepared under three different growth conditions. These samples were observed using the POSAP in order to study both the composition of the wells and the interface morphology. The first set of wells examined were prepared in a conventional reactor to which a quartz wool baffle had been added to promote gas intermixing. The effect of this was to hold a volume of gas within the chamber between growth stages, leading to a structure where the wells had a composition of GalnAsP lattice matched to the InP barriers, and where the interfaces were very indistinct. A POSAP image showing a well in this sample is shown in figure 1. The second set of wells were grown in the same reactor but with the quartz wool baffle removed. This set of wells were much better defined, as can be seen in figure 2, and the wells were much closer to the intended composition, but still with measurable levels of phosphorus. The final set of wells examined were prepared in a reactor where the design had the effect of minimizing the recirculating volume of gas. In this case there was again further improvement in the well quality. It also appears that the left hand side of the well in figure 2 is more abrupt than the right hand side, indicating that the switchover at this interface from barrier to well growth is more abrupt than the switchover at the other interface.


1988 ◽  
Vol 53 (3) ◽  
pp. 316-327 ◽  
Author(s):  
Alan G. Kamhi ◽  
Hugh W. Catts ◽  
Daria Mauer ◽  
Kenn Apel ◽  
Betholyn F. Gentry

In the present study, we further examined (see Kamhi & Catts, 1986) the phonological processing abilities of language-impaired (LI) and reading-impaired (RI) children. We also evaluated these children's ability to process spatial information. Subjects were 10 LI, 10 RI, and 10 normal children between the ages of 6:8 and 8:10 years. Each subject was administered eight tasks: four word repetition tasks (monosyllabic, monosyllabic presented in noise, three-item, and multisyllabic), rapid naming, syllable segmentation, paper folding, and form completion. The normal children performed significantly better than both the LI and RI children on all but two tasks: syllable segmentation and repeating words presented in noise. The LI and RI children performed comparably on every task with the exception of the multisyllabic word repetition task. These findings were consistent with those from our previous study (Kamhi & Catts, 1986). The similarities and differences between LI and RI children are discussed.


2002 ◽  
Vol 16 (2) ◽  
pp. 114-118 ◽  
Author(s):  
Timo Ruusuvirta ◽  
Heikki Hämäläinen

Abstract Human event-related potentials (ERPs) to a tone continuously alternating between its two spatial loci of origin (middle-standards, left-standards), to repetitions of left-standards (oddball-deviants), and to the tones originally representing these repetitions presented alone (alone-deviants) were recorded in free-field conditions. During the recordings (Fz, Cz, Pz, M1, and M2 referenced to nose), the subjects watched a silent movie. Oddball-deviants elicited a spatially diffuse two-peaked deflection of positive polarity. It differed from a deflection elicited by left-standards and commenced earlier than a prominent deflection of negative polarity (N1) elicited by alone-deviants. The results are discussed in the context of the mismatch negativity (MMN) and previous findings of dissociation between spatial and non-spatial information in auditory working memory.


Sign in / Sign up

Export Citation Format

Share Document