With the continuous growth of dimensions in retrieval systems, only a few data points are distributed near the center (empty space phenomenon), and the distance between data points in high-dimensional space is nearly equal (dimensional effect), resulting in high complexity and low accuracy in retrieval. Aiming at the preceding problems, this article designs a speech secure hash retrieval scheme. In this scheme, the spectral subband centroids of speech are extracted to generate the feature vector, then the biometric template index is established by KDTree classification, and the specific SHA256-Ushiki chaotic encryption algorithm key is allocated to each index. The security framework is constructed according to the cancelable biometric template generated by the combination of classification and distribution key, and the binary hash vector is generated, then the hash vector is encrypted. Experimental results show that through the establishment of the KDTree cancelable biometric template index, the super rectangular region of the K -dimensional space is constructed, which effectively solves the empty space phenomenon and the dimensional effect. Through the KDTree nearest neighbor search, the algorithm reduces the number of matches between classes, which effectively reduces computational complexity and accuracy problems. The tampering comparison of mobile terminal realizes the content verifiable retrieval. The speech encryption effectively prevents the leakage of plaintext and ensures security of the speech storage and transmission process.

Hashing has received a lot of attention in large-scale image retrieval due to its high retrieval accuracy and speed. Unsupervised deep hashing methods with pseudo-labels suffer from suboptimal performance due to low clustering accuracies, which result in unreliable generated pseudo-labels, and are highly sensitive to the number of clusters. To tackle these challenges, we propose an unsupervised deep hashing image retrieval method with dynamic pseudo-multi-labels (UDHPM). Specifically, UDHPM designs a dynamic pseudo-multi-label generation network by employing a soft clustering model to maximize the approximation to the real data distribution and continuously optimizing the pseudo-multi-labels in an end-to-end manner to provide reliable supervised information. Furthermore, UDHPM preserves the category information of dynamic pseudo-multilabels by applying Kullback–Leibler (KL) divergence to the hashing network. In experiments on three benchmark datasets, UDHPM significantly improves retrieval performance over existing state-of-the-art unsupervised deep hashing methods. UDHPM also exhibits a degree of robustness to the number of clusters.

Deep hashing enhances image retrieval accuracy by integrating hash encoding with deep neural networks. However, existing unsupervised deep hashing methods primarily rely on the rotational invariance of images to construct triplets, resulting in triplets that are unsatisfactory in both reliability and quantity. Additionally, some methods fail to adequately consider the relative similarity information between samples. To overcome these limitations, we propose a novel unsupervised deep triplet hashing method for image retrieval (abbreviated as UDTrHash). UDTrHash utilizes the extremal cosine similarity of deep features of images to construct more reliable first type triplets and expands the formed triplets through data augmentation strategies to introduce a larger number of triplets. Furthermore, we design a new triplet loss function to enhance the discriminative ability of the generated hash codes. Extensive experiments demonstrate that UDTrHash exhibits superior performance on three public benchmark datasets such as MIRFlickr25K compared to existing state-of-the-art hashing methods.