언어 모델 기반 음성 특징 추출을 활용한 생성 음성 탐지
Voice Synthesis Detection Using Language Model-Based Speech Feature Extraction
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
최근 음성 생성 기술의 급격한 발전으로, 텍스트만으로도 자연스러운 음성 합성이 가능해졌다. 이러한 발전은 타인의 음성을 생성하여 범죄에 이용하는 보이스피싱과 같은 악용 사례를 증가시키는 결과를 낳고 있다. 음성 생성 여부를 탐지하는 모델은 많이 개발되고 있으며, 일반적으로 음성의 특징을 추출하고 이러한 특징을 기반으로 음성 생성 여부를 탐지한다. 본 논문은 생성 음성으로 인한 악용 사례에 대응하기 위해 새로운 음성 특징 추출 모델을 제안한다. 오디오를 입력으로 받는 딥러닝 기반 오디오 코덱 모델과 사전 학습된 자연어 처리 모델인 BERT를 사용하여 새로운 음성 특징 추출 모델을 제안하였다. 본 논문이 제안한 음성 특징 추출 모델이 음성 탐지에 적합한지 확인하기 위해 추출된 특징을 활용하여 4가지 생성 음성 탐지 모델을 만들어 성능평가를 진행하였다. 성능 비교를 위해 기존 논문에서 제안한 Deepfeature 기반의 음성 탐지 모델 3개와 그 외 모델과 정확도 및 EER을 비교하였다. 제안한 모델은 88.08%로 기존 모델보다 높은 정확도와 11.79%의 낮은 EER을 보였다. 이를 통해 본 논문에서 제안한 음성 특징 추출 방법이 생성 음성과 실제 음성을 판별하는 효과적인 도구로 사용될 수 있음을 확인하였다.
Abstract
Recent rapid advancements in voice generation technology have enabled the natural synthesis of voices using text alone. However, this progress has led to an increase in malicious activities, such as voice phishing (voishing), where generated voices are exploited for criminal purposes. Numerous models have been developed to detect the presence of synthesized voices, typically by extracting features from the voice and using these features to determine the likelihood of voice generation.This paper proposes a new model for extracting voice features to address misuse cases arising from generated voices. It utilizes a deep learning-based audio codec model and the pre-trained natural language processing model BERT to extract novel voice features. To assess the suitability of the proposed voice feature extraction model for voice detection, four generated voice detection models were created using the extracted features, and performance evaluations were conducted. For performance comparison, three voice detection models based on Deepfeature proposed in previous studies were evaluated against other models in terms of accuracy and EER. The model proposed in this paper achieved an accuracy of 88.08% and a low EER of 11.79%, outperforming the existing models. These results confirm that the voice feature extraction method introduced in this paper can be an effective tool for distinguishing between generated and real voices.
참고문헌 (20)
- "Asvspoof 2019: Automatic speaker verification spoofing and countermeasures challenge evaluation plan" (2020)[URL]
- Yi, Jiangyan, "Add 2022: the first audio deep synthesis detection challenge", ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2022), IEEE
- Xiao, Xiong, "Spoofing speech detection using high dimensional magnitude and phase features : the NTU approach for ASVspoof 2015challenge", Interspeech (2015)
- Chen, Lian-Wu, "Speaker verification against synthetic speech", 2010 7th International Symposium on Chinese Spoken Language Processing (2010), IEEE
- Das, Rohan Kumar, "Long range acoustic and deep features perspective on ASVspoof 2019", 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) (2019), IEEE
- M. Sahidullah, "A comparisonof features for synthetic speech detection", Proc. of INTER-SPEECH (2015)
- T. B. Patel, "Combining evidences from mel cepstral,cochlear filter cepstral and instantaneous frequency features fordetection of natural vs. spoofed speech", Conference of Internaꠓtional Speech Communication Association (2015)
- T. K. A, "An overview of text-independent speakerrecognition : From features to supervectors", Speech Communication, 52(1), pp.12~40 (2010)
- Y. Xie, "Siamese network with wav2vecfeature for spoofing speech detection", Interspeech (2021)
- X. Wang, "Investigating self-supervised front ends for speech spoofing countermeasures", The Speaker and Language Recognition Workshop (2021)
- Zeghidour, Neil, "Soundstream:An end-to-end neural audio codec", IEEE, 2021, pp.30, ~
- Kim Seung-min, "Speech Generation Classifier Using BERT", Korea Institute of Information Security & Cryptology (2023)
- He, Kaiming, "Deep residual learning for image recognition", Proceedings of the IEEE conference on computer vision and pattern recognition (2016)
- Yadav, Amit Kumar Singh, "ASSD: Synthetic Speech Detection in the AAC Compressed Domain", ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2023), IEEE
- G. Hua, "Towards End-to-End Synthetic Speech Detection", IEEE Signal Processing Letters, 28, pp.1265~1269 (2021)
- E. R. Bartusiak, "Synthesized Speech Detection Using Convolutional Transformer-Based Spectrogram Analysis", Proceedings of the IEEE Asilomar Conference on Signals, Systems, and Computers, pp.1426~1430 (2021)