Search

Have you ever wondered how a song might sound if performed by a different artist? In this work, we propose SCM-GAN, an end-to-end non-parallel song conversion system powered by generative adversarial and transfer learning that allows users to listen to a selected target singer singing any song. SCM-GAN first separates songs into vocals and instrumental music using a U-Net network, then converts the vocal segments to the target singer using advanced CycleGAN-VC, before merging the converted vocals with their corresponding background music. SCM-GAN is first initialized with feature representations learned from a state-of-the-art voice-to-voice conversion and then trained on a dataset of non-parallel songs. Furthermore, SCM-GAN is evaluated against a set of metrics including global variance GV and modulation spectra MS on the 24 Mel-cepstral coefficients (MCEPs). Transfer learning improves the GV by 35% and the MS by 13% on average. A subjective comparison is conducted to test the user satisfaction with the quality and the naturalness of the conversion. Results show above par similarity between SCM-GAN's output and the target (70\% on average) as well as great naturalness of the converted songs.

LaRa (Lander Radioscience) is an experiment on the ExoMars 2020 mission that uses the Doppler shift on the radio link due to the motion of the ExoMars platform tied to the surface of Mars with respect to the Earth ground stations (e.g. the deep space network stations of NASA), in order to precisely measure the relative velocity of the lander on Mars with respect to the Earth. The LaRa measurements shall improve the understanding of the structure and processes in the deep interior of Mars by obtaining the rotation and orientation of Mars with a better precision compared to the previous missions. In this paper, we provide the analysis done until now for the best realization of these objectives. We explain the geophysical observation that will be reached with LaRa (Length-of-day variations, precession, nutation, and possibly polar motion). We develop the experiment set up, which includes the ground stations on Earth (so-called ground segment). We describe the instrument, i.e. the transponder and its three antennas. We further detail the link budget and the expected noise level that will be reached. Finally, we detail the expected results, which encompasses the explanation of how we shall determine Mars' orientation parameters, and the way we shall deduce Mars' interior structure and Mars' atmosphere from them. Lastly, we explain briefly how we will be able to determine the Surface platform position., Comment: 43 pages, 39 figures, accepted in Planetary and Space Science

Energy dispersion and spin orientation of the protected states at interfaces between topological insulators (TIs) and non-topological materials depend on the charge redistribution, strain, and atomic displacement at the interface. Knowledge of these properties is essential for applications of topological compounds, but direct access to them in the interface geometry is difficult. We show that conductance of a gated double junction at the surface of a topological insulator exhibits oscillations and a quasi-linear decay as a function of gate voltage in different regimes. These give the values for the quasiparticle velocities along and normal to the junction in the interface region, and determine the symmetry of the topological interface states. The results are insensitive to the boundary conditions at the junction.

Monocular Odometry systems can be broadly categorized as being either Direct, Indirect, or a hybrid of both. While Indirect systems process an alternative image representation to compute geometric residuals, Direct methods process the image pixels directly to generate photometric residuals. Both paradigms have distinct but often complementary properties. This paper presents a Unified Formulation for Visual Odometry, referred to as UFVO, with the following key contributions: (1) a tight coupling of photometric (Direct) and geometric (Indirect) measurements using a joint multi-objective optimization, (2) the use of a utility function as a decision maker that incorporates prior knowledge on both paradigms, (3) descriptor sharing, where a feature can have more than one type of descriptor and its different descriptors are used for tracking and mapping, (4) the depth estimation of both corner features and pixel features within the same map using an inverse depth parametrization, and (5) a corner and pixel selection strategy that extracts both types of information, while promoting a uniform distribution over the image domain. Experiments show that our proposed system can handle large inter-frame motions, inherits the sub-pixel accuracy of direct methods, can run efficiently in real-time, can generate an Indirect map representation at a marginal computational cost when compared to traditional Indirect systems, all while outperforming state of the art in Direct, Indirect and hybrid systems.

The bulk Rashba semiconductors BiTeX (X=I, Cl and Br) with intrinsically enhanced Rashba spin-orbit coupling provide a new platform for investigation of spintronic and magnetic phenomena in materials. We theoretically investigate the interlayer exchange interaction between two ferromagnets deposited on opposite surfaces of a bulk Rashba semiconductor BiTeI in its trivial and topological insulator phases. In the trivial phase BiTeI, we find that for ferromagnets with a magnetization orthogonal to the interface, the exchange coupling is reminiscent of that of a conventional three-dimensional metal. Remarkably, ferromagnets with a magnetization parallel to the interface display a magnetic exchange qualitatively different from that of conventional three-dimensional metal due to the spin-orbit coupling. In this case, the interlayer exchange interaction acquires two periods of oscillations and decays as the inverse of the thickness of the BiTeI layer. For topological BiTeI, the magnetic exchange interaction becomes mediated only by the helical surface states and acts between the one-dimensional spin chains at the edges of the sample. The surface state-mediated interlayer exchange interaction allows for the coupling of ferromagnets with non-collinear magnetization and displays a decay power different from that of trivial BiTeI, allowing the detection of the topological phase transition in this material. Our work provides insights into the magnetic properties of these newly discovered materials and their possible functionalization.

There is a long history of radio telescopes being used to augment the radio antennas regularly used to conduct telemetry, tracking, and command of deep space spacecraft. Radio telescopes are particularly valuable during short-duration mission critical events, such as planetary landings, or when a mission lifetime itself is short, such as a probe into a giant planet's atmosphere. By virtue of its high sensitivity and frequency coverage, the next-generation Very Large Array would be a powerful addition to regular spacecraft ground systems. Further, the science focus of many of these deep-space missions provides a "ground truth" in the solar system that complements other aspects of the ngVLA's science case, such as the formation of planets in proto-planetary disks., Comment: 6 pages, 1 figure; To be published in the ASP Monograph Series, "Science with a Next-Generation VLA", ed. E. J. Murphy (ASP, San Francisco, CA)

Depositing monolayer graphene on a transition metal dichalcogenide (TMD) semiconductor substrate has been shown to change the dynamics of the electronic states in graphene, inducing spin orbit coupling (SOC) and staggered potential effects. Theoretical studies on commensurate supercells have demonstrated the appearance of interesting phases, as different materials and relative gate voltages are applied. Here we address the effects of the real incommensurability between lattices by implementing a continuum model approach that does not require small-period supercells. The approach allows us to study the role of possible relative twists of the layers, and verify that the SOC transfer is robust to twists, in agreement with observations. We characterize the nature of the different phases by studying an effective Hamiltonian that fully describes the graphene-TMD heterostructure. We find the system supports topologically non-trivial phases over a wide range of parameter ranges, which require the dominance of the intrinsic SOC over the staggered and Rashba potentials. This tantalizing result suggests the possible experimental realization of a tunable quantum spin Hall phase under suitable conditions. We estimate that most TMDs used to date likely result in weak intrinsic SOC that would not drive the heterostructure into topologically non-trivial phases. Additional means to induce a larger intrinsic SOC, such as strain fields or heavy metal intercalation may be required., Comment: 7 pages, 4 figures

We study how non-helical spin textures at the boundary between a topological insulator (TI) and a superconductor (SC) affect the proximity-induced superconductivity of the TI interface state. We consider TIs coupled to both spin-singlet and spin-triplet SCs, and show that for the spin-triplet parent SCs the resulting order parameter induced onto the interface state sensitively depends on the symmetries which are broken at the TI-SC boundary. For chiral spin-triplet parent SCs, we find that nodal proximity-induced superconductivity emerges when there is broken twofold rotational symmetry which forces the spins of the non-helical topological states to tilt away from the interface plane. We furthermore show that the Andreev conductance of lateral heterostructures joining TI-vacuum and TI-SC interfaces yields experimental signatures of the reduced symmetries of the interface states., Comment: 5 pages, 2 figures

Visual Odometry (VO) can be categorized as being either direct or feature based. When the system is calibrated photometrically, and images are captured at high rates, direct methods have shown to outperform feature-based ones in terms of accuracy and processing time; they are also more robust to failure in feature-deprived environments. On the downside, Direct methods rely on heuristic motion models to seed the estimation of camera motion between frames; in the event that these models are violated (e.g., erratic motion), Direct methods easily fail. This paper proposes a novel system entitled FDMO (Feature assisted Direct Monocular Odometry), which complements the advantages of both direct and featured based techniques. FDMO bootstraps indirect feature tracking upon the sub-pixel accurate localized direct keyframes only when failure modes (e.g., large baselines) of direct tracking occur. Control returns back to direct odometry when these conditions are no longer violated. Efficiencies are introduced to help FDMO perform in real time. FDMO shows significant drift (alignment, rotation & scale) reduction when compared to DSO & ORB SLAM when evaluated using the TumMono and EuroC datasets.