| With the advent of big data and cloud computing era,there is a large amount of audio data in the cloud server.In the process of audio retrieval in the semi-open cloud environment,audio data and user information are easy to leak.At the same time,the current audio hashing retrieval algorithm has some problems,such as single hashing construction process,which can not effectively protect the audio features.To solve the above problems,this thesis takes audio hash retrieval algorithm as the main line,combined with feature security template protection technology,chaos mapping encryption and other technologies to carry out research.The main work is as follows:1.In order to improve the protection of audio features and the diversity of biological hash construction process,the existing audio hash retrieval algorithm can solve the problem that the hash construction process is single and cannot form effective protection for audio features.An audio hash retrieval algorithm based on Lu’s feature security template is proposed.Firstly,the audio features of the original audio were extracted by GTCC algorithm and dimensionality reduction was carried out on the audio features.Then,the feature vectors after dimensionality reduction were differential processed and classified by SVM.The key generated from the classification results was used as the initial value of Lu’s chaotic map to construct the Lu’s feature security template.Finally,the eigenvectors after dimensionality reduction are quantized by Lu’s security template to obtain biological hashes.The experimental results show that the proposed safety template improves the security and diversity of the biological hash construction process.In addition,the audio biological hashes produced by Lu’s characteristic security template have better performance.2.For feature security template protection,there are problems such as the singleness of template generation in the audio hash retrieval process and the vulnerability of template leakage due to key management.An audio hash retrieval algorithm based on VAD-SCC feature security template protection is proposed.Firstly,the client performs similarity retrieval according to the VAD audio template to complete the audio template matching.Then,based on the detection result of the template and the independent key provided by the user as the initial value of the cosine chaotic system,the corresponding VAD-SCC feature security template is constructed through multiple iterations.Finally,the audio spectral features after dimensionality reduction are quantized by the feature security template,so as to obtain the biological hash sequence.The experimental results show that the proposed VAD-SCC feature security template has good security and matching performance,and can keep the diversity and revocation of the feature template.At the same time,the audio hash retrieval algorithm has higher discrimination and robustness,which can meet the need of multiple content retention operation audio retrieval.In this thesis,the audio hash retrieval algorithm based on feature security template is studied and some scientific research results are obtained.The research shows that this scheme can protect the audio feature data and has good retrieval performance.In addition,the research direction of the next stage of audio hash retrieval is proposed,including how to further improve the efficiency of audio hash retrieval and how to apply it in practical application scenarios. |
