Generating human motion in complex 3D scenes from text is a challenging task with broad applications. However, existing methods often overlook realistic physical contact, resulting in visually plausible but physically unrealistic motion, e.g., penetration. To alleviate this, we propose IntentMotion, a novel framework that generates human motion in 3D scenes from natural language instructions by explicitly modeling intent. We first introduce the Intention-Guided Contact Field (IGCF). This differentiable voxel-based contact region representation explicitly aligns parsed language roles with spatial contact regions through a hierarchical attention mechanism. IGCF is jointly trained with a diffusion-based motion generator, allowing contact predictions to adapt dynamically through gradient feedback. To improve the controllability and physics-aware motion, we further propose an Intention-Aware Diffusion Model (IADM), which decouples the high-level semantic planning from the low-level contact refinement in a coarse-to-fine process. The optimized contact cues are utilized to guide the synthesis of a coarse trajectory, followed by refining detailed pose sequences under IGCF supervision. Experiments on the HUMANISE and LINGO datasets demonstrate that our IntentMotion outperforms recent baselines in contact accuracy, semantic alignment, and generalization to unseen scenes.

It is a challenging task to create realistic 3D avatars that accurately replicate individuals' speech and unique talking styles for speech-driven facial animation. Existing techniques have made remarkable progress but still struggle to achieve lifelike mimicry. This article proposes "TalkingStyle", a novel method to generate personalized talking avatars while retaining the talking style of the person. Our approach uses a set of audio and animation samples from an individual to create new facial animations that closely resemble their specific talking style, synchronized with speech. We disentangle the style codes from the motion patterns, allowing our method to associate a distinct identifier with each person. To manage each aspect effectively, we employ three separate encoders for style, speech, and motion, ensuring the preservation of the original style while maintaining consistent motion in our stylized talking avatars. Additionally, we propose a new style-conditioned transformer decoder, offering greater flexibility and control over the facial avatar styles. We comprehensively evaluate TalkingStyle through qualitative and quantitative assessments, as well as user studies demonstrating its superior realism and lip synchronization accuracy compared to current state-of-the-art methods.

This research aims at the robust and high-resolution reconstruction of asteroid’s interior structure using lander-to-orbiter radar system. In recent years, the full waveform inversion (FWI) has been suggested as a potential method to asteroid tomography. Due to the limitation of computing capacity, FWI is usually performed by local rather than global optimization method, which makes it suffer from local minima problem especially when the signal lacks low-frequency components. To ensure the global convergence, FWI requires the initial model be accurate enough to avoid the local minima. But in practice, the low-frequency components are naturally absent in the remote radar signal as the limitation of bandwidth, and the prior information of asteroid is usually not sufficient to build an accurate initial model, which leads that using conventional FWI directly may not be capable to obtain a robust reconstruction. Considering the above problems, envelope inversion (EI) which works on the baseband signal and therefore, can recover the long-wavelength structure of asteroid is proposed as a supplementary to FWI. Initial model dependence and noise sensitivity of FWI and EI in asteroid tomography are analyzed based on 2-D numerical experiments. The EI + FWI combination constrained by total variation regularization shows the characteristics of good independence on the initial model and high imaging resolution. Based on EI + FWI strategy, a series of 3-D numerical experiments are conducted to test the influence of orbital measurement density and landing site on the tomography.

The orthogonal frequency division multiplexing (OFDM) based ground-penetrating radar (GPR) has been proven to have many advantages. The estimation of amplitudes and time delays of interface echoes is significant particularly in GPR systems for pavement profiling. For the recently developed OFDM-based GPR systems, this letter proposes an algorithm for precisely estimating the echo amplitudes and time delays. We estimate the propagation channel in frequency domain and transform the resulted channel frequency response to time domain. Considering the reduction of the channel impulse response (CIR) leakage caused by IFFT, a windowing method is applied before the IFFT. The challenge is that the general windowing method leads to resolution decrease and peak value losses for the windowed CIR, which contains the echo amplitude and time delay information. We design a window that causes very little echo amplitude loss and improves the estimation resolution. In this way, the echo amplitudes and time delays can be obtained from the largest few extreme values of the windowed CIR. The algorithm performs a low complexity, and the simulation results show that the estimation errors for both amplitudes and time delays are small with the range resolution improved.