In recent years, text classification has found wide application in diverse real-world scenarios. In Chinese news classification tasks, limitations such as sparse contextual information and semantic ambiguity exist in the title text. To improve the performance of short text classification, this paper proposes a Word2Vec-based enhanced word embedding method and exhibits the design of a dual-channel hybrid neural network architecture to effectively extract semantic features. Specifically, we introduce a novel weighting scheme, Term Frequency-Document Frequency Category-Distribution Weight (TF-IDF-CDW), where Category Distribution Weight (CDW) reflects the distribution pattern of words across different categories. By weighting the pretrained Word2Vec vectors with TF-IDF-CDW and concatenating them with part-of-speech (POS) feature vectors, semantically enriched and more discriminative word embedding vectors are generated. Furthermore, we propose a dual-channel hybrid model based on a Gated Convolutional Neural Network (GCNN) and Bidirectional Long Short-Term Memory (BiLSTM), which jointly captures local features and long-range global dependencies. To evaluate the overall performance of the model, experiments were conducted on the Chinese short text datasets THUCNews and TNews. The proposed model achieved classification accuracies of 91.85% and 87.70%, respectively, outperforming several comparative models and demonstrating the effectiveness of the proposed method.
In gesture recognition based on millimeter-wave radar, generating spectrograms is typically independent of the actual application and designed separately. In this case, the task is simply decoupled, resulting in the generated spectrograms from radar signals not being optimally suited for the recognition task. Additionally, the emergence of gesture categories representing new semantics requires the recollection of a large amount of high-quality labeled data and retraining of the model. To address these problems, we propose a radar-based category-scalable gesture recognition framework, R-CSGR, for gesture spectrogram generation and two-stage gesture recognition. Considering the noise and environmental factors, only gesture-related signals are extracted and aggregated in the doppler and angle dimensions to form a location-independent, information-dense gesture spectrogram for the two-stage recognition. In the first stage, the reconstruction of spectrogram for the original categories is used as a self-supervised learning task to utilize low-cost unlabeled data. In the second stage, the Classification Layer based on the cosine nearest-centroid method is used to quickly recognize new gesture categories whereas maintaining the recognition capability of the original categories. The result shows that with the introduction of five new gesture categories and only eight shots per category in the support set, an average recognition accuracy of 96.88% is achieved for all nine gesture categories.