Audio samples from "SecureSpeech: Prompt-based Speaker and Content Protection"

Paper:

Accepted by IJCB 2025

Authors:

Messages about Data:

• Audio samples are gathered from SLUE-VoxPopuli. The SLUE-VoxPopuli dataset contains a subset of VoxPopuli dataset and the copyright of this subset remains the same with the original license, CC0.

   

  ---

   

  Abstract of "SecureSpeech: Prompt-based Speaker and Content Protection":

  Given the increasing privacy concerns from identity theft and the re-identification of speakers through content in the speech field, it is crucial to develop solutions that ensure both speaker identity protection and content preservation. This paper proposes a prompt-based speech generation pipeline that ensures dual anonymization of both speaker identity and spoken content. This is addressed through 1) generating a speaker identity unlinkable to source speaker, controlled by descriptors, and 2) replacing of sensitive content within original text using a name entity recognition model and a large language model. The pipeline utilizes the anonymized speaker identity and text to generate high-fidelity, privacy-friendly speech via a text-to-speech synthesis model. Experimental results demonstrates an achievement of significant privacy protection while maintaining a decent level of content retention and audio quality. This paper also investigates the impact of varying speaker descriptions on utility and privacy of generated speech to determine if potential biases.

   

  Demo: Random Speaker Speech Audioset from VoxPopuli and Synthetic Audioset Generated

  	Original Speech: Audio	Original Speech: Transcript	NER Instructions	Synthetic Speech: Speaker Description	Synthetic Speech: Audio	Synthetic Speech: Transcript
  20100518-0900-PLENARY-8-en_20100518-12:17:06_5		involved for their positive and constructive approach.	"{'type': array([], dtype=object), 'start': array([], dtype=int32), 'length': array([], dtype=int32)}"	A male voice in a Finnish accent reads a book normally with a low-pitched voice. The speaker's voice is clean		involved for their positive and constructive approach.
  20090311-0900-PLENARY-6-en_20090311-10:26:44_11		my point is the following in a few weeks you will meet mr obama the new president of the united states	{'type': array(['PLACE', 'PERSON', 'WHEN'], dtype=object), 'start': array([89, 58, 26], dtype=int32), 'length': array([13, 5, 14], dtype=int32)}	A male voice in a South African accent reads a book very slowly with a low-pitched and monotone voice. The speaker's voice is clean		My point is the following in a few months you will meet Mr. Smith the new president of the Asian federation.
  20131022-0900-PLENARY-19-en_20131022-20:29:44_1		i welcome very much the resumption of talks between the israelis and the palestinians and sincerely hope that they will succeed.	"{'type': array(['NORP', 'NORP'], dtype=object), 'start': array([73, 56], dtype=int32), 'length': array([12, 8], dtype=int32)}"	A male voice in a Finnish accent reads a book very slowly with a high-pitched and monotone voice. The recording is excellent		i welcome very much the resumption of talks between the citizens and the residents and sincerely hope that they will succeed.

   

  Demo: Specific Speaker Speech Audioset from VoxPopuli and Synthetic Audioset Generated

  	Original Speech: Audio	Original Speech: Transcript	NER Instructions	Synthetic Speech: Speaker Description	Synthetic Speech: Audio	Synthetic Speech: Transcript
  20090218-0900-PLENARY-10-en_20090218-15:15:21_13		too often the region has been plagued by cycles of violence rising extremism and economic hardship. at the same time the conditions for europeans and americans to work together in the search for peace in the middle east are probably	"{'type': array(['PLACE', 'NORP', 'NORP'], dtype=object), 'start': array([208, 150, 136], dtype=int32), 'length': array([11, 9, 9], dtype=int32)}"	Joy reads a book with a low-pitched and expressive and animated voice normally.		Too often the region has been plagued by cycles of violence rising extremism and economic hardship. At the same time the conditions for Asians and Canadians to work together in the search for peace in the southern region are probably.
  20110927-0900-PLENARY-15-en_20110927-21:02:59_11		this behaviour is unacceptable.	"{'type': array([], dtype=object), 'start': array([], dtype=int32), 'length': array([], dtype=int32)}"	Will reads a book with a high-pitched voice quickly. The speaker's voice is clean		this behaviour is unacceptable.
  20130114-0900-PLENARY-17-en_20130114-22:29:53_6		i call for deeper knowledge of the most recent history of europe.	"{'type': array(['PLACE'], dtype=object), 'start': array([58], dtype=int32), 'length': array([6], dtype=int32)}"	Gary reads a book with a high-pitched and expressive and animated voice very slowly. The speaker's voice is clean		I call for deeper knowledge of the most recent history of Asia.

   

   

  ---