Pointing at the vexed question of blind recognition in the convolutional code class, this paper proposes a convolutional code blind identification method via convolutional neural networks (CNNs). First, this algorithm uses the traditional method to generate different convolutional codes, and the feature extraction algorithm adopts the theorem of Euclid’s algorithm. Then, the input signal is loaded to the CNN; next, the feature is extracted by convolutional kernel. Finally, the Softmax activation function is applied to full‐connection layer network. After the input signals pass through the above layers, the system classifies the signals. The research results indicate that the presented algorithm has improved the recognition performance of code length and rate. For different convolutional codes with parameters of (5, 7), (15, 17), (23, 35), (53, 75), and (133, 171) and similar convolutional codes with parameters of (3, 1, 6), (3, 1, 7), (2, 1, 7), (2, 1, 6), and (2, 1, 5), the recognition rate of parameter classification can reach 100% at signal‐to‐noise ratio (SNR) of 3 dB.

As a selective state space model, Mamba exhibits outstanding performance and efficiency in sequence modeling tasks. Therefore, in this paper, we use Mamba as the fundamental network component to construct a novel speech separation model, Speech Conv-Mamba. Specifically, this model embeds Mamba within a U-shaped convolutional network to build the encoder and decoder network for high-dimensional representation and waveform reconstruction of speech signals. Additionally, we stack multiple temporal dilated convolutions and Mamba to create the separation network for separation task. Our comparative experiments on the GRID2Mix and Libri2Mix datasets demonstrate that the proposed model Speech Conv-Mamba, which achieves 98% and 89% of SepFormer's separation accuracy on two datasets using only 9% (2.4 M) of its model size, provides much less computational complexity and training cost.

Audio-only speech separation methods cannot fully exploit audio-visual correlation information of speaker, which limits separation performance. Additionally, audio-visual separation methods usually adopt traditional idea of feature splicing and linear mapping to fuse audio-visual features, this approach requires us to think more about fusion process. Therefore, in this paper, combining with the changes of speaker mouth landmarks, we propose a time-domain audio-visual temporal convolution attention speech separation method (AVTA). In AVTA, we design a multiscale temporal convolutional attention (MTCA) to better focus on contextual dependencies of time sequences. We then use sequence learning and fusion network composed of MTCA to build a separation model for speech separation task. On different datasets, AVTA achieves competitive performance, and compared to baseline methods, AVTA is better balanced in training cost, computational complexity and separation performance.

The aerial light field (ALF) can render higher quality images of large-scale 3D scenes. In this paper, we apply the ALF technology to study the image captured and novel view rendering of unmanned aerial vehicle (UAV), which exists some problems, such as large scene and depth of field. First, we design an ALF sampling model using spectral analysis of Fourier theory. Based on the ALF sampling model, the exact expression of ALF spectrum is derived. By the spectral support of ALF, we analyze the influence of pitch angle on light field sampling and its bandwidth. Particularly, the bandwidth of ALF can be applied to determine the minimum sampling rate for UAV. Additionally, we design a reconstruction filter that is related to pitch angle to render novel views of UAV. Finally, our experiments show that our sampling and rendering methods can improve the rendering quality of UAV novel view rendering.

In the previous results, light field sampling is based on ideal assumptions (e.g., Lambertian and Non-occluded scene), and thus we would like to more precisely analyze the sparsity sampling of light field signal. We present a sparsity analysis of light field (SALF) method for optimizing light field sampling rate. The SALF method applies the Fourier projection-slice theorem to simplify the initialization of light field sampling. Furthermore, we use a voting scheme to select light field spectra in which the frequency coefficients are nonzero. These spectra include many scene information and their captured positions are approximately equal to camera positions in the frequency domain. If the camera is only placed in these selected camera positions, the sampling rate can be optimized and the rendering quality can be guaranteed. Finally, we compare SALF method with other light field sampling methods to verify the claimed performance. The reconstruction results show that the SALF method improves rendering quality of novel views and outperforms those of other comparison methods.

A number of theories have been proposed for reducing sampling rate of light field. But these theories still need a great many of samples (images) to obtain sufficient geometric information. In this paper, we utilize the sparse representation of light field in Discrete Cosine Transform domain to present a Discrete Cosine Sparse Basis (DCSB). Thus, we can find out the zeros of DCSB to reduce sampling requirement of light field for alias-free rendering. Finally, experimental results demonstrate the effectiveness of our approach without lose information.