Uploaded August 2026 | Updated September 2026, 2 weeks ago
DNA data storage has emerged as a promising medium for large-scale, long-term archival preservation. However, existing studies have primarily focused on text, images, and general digital files, while dedicated systems for speech data remain largely unexplored. To fill this gap, we propose VoiceArk, a progressive DNA storage system for robust and scalable speech archiving. As the first stage of the system, we develop a parameterized base layer that employs speech-oriented compression to convert waveform speech into a low-bitrate parameter stream, thereby enabling high compression efficiency, improved error robustness, and intelligible speech reconstruction. To further support progressive access, we envision a deep learning- based enhancement layer that can incrementally improve reconstruction quality as additional sequencing reads become available. Under this design, the system is able to first recover intelligible speech at low read cost and then progressively refine perceptual quality with increased read allocation. Experiments conducted in a simulated DNA channel error environment show that the parameterized base layer achieves intelligible speech reconstruction without error-correcting codes, reaching an STOI score of 0.7 at an error rate of 0.1%. These results demonstrate that the proposed base layer provides a robust speech representation for DNA storage and establishes a practical foundation for progressive speech archiving under a flexible trade-off between read cost and reconstruction quality.
Presented by
Tao Han, Tianjin University
This is a presentation from the 2026 Storage and Computing with DNA Conference.
· Learn More about the SNIA DNA Data Storage Alliance: snia.org/groups/snia-dna-technology-affiliate
· SNIA Educational Library: snia.org/library
· X: twitter.com/SNIA
· LinkedIn: linkedin.com/company/snia
DNA data storage has emerged as a promising medium for large-scale, long-term archival preservation. However, existing studies have primarily focused on text, images, and general digital files, while dedicated systems for speech data remain largely unexplored. To fill this gap, we propose VoiceArk, a progressive DNA storage system for robust and scalable speech archiving. As the first stage of the system, we develop a parameterized base layer that employs speech-oriented compression to convert waveform speech into a low-bitrate parameter stream, thereby enabling high compression efficiency, improved error robustness, and intelligible speech reconstruction. To further support progressive access, we envision a deep learning- based enhancement layer that can incrementally improve reconstruction quality as additional sequencing reads become available. Under this design, the system is able to first recover intelligible speech at low read cost and then progressively refine perceptual quality with increased read allocation. Experiments conducted in a simulated DNA channel error environment show that the parameterized base layer achieves intelligible speech reconstruction without error-correcting codes, reaching an STOI score of 0.7 at an error rate of 0.1%. These results demonstrate that the proposed base layer provides a robust speech representation for DNA storage and establishes a practical foundation for progressive speech archiving under a flexible trade-off between read cost and reconstruction quality.
Presented by
Tao Han, Tianjin University
This is a presentation from the 2026 Storage and Computing with DNA Conference.
· Learn More about the SNIA DNA Data Storage Alliance: snia.org/groups/snia-dna-technology-affiliate
· SNIA Educational Library: snia.org/library
· X: twitter.com/SNIA
· LinkedIn: linkedin.com/company/snia
