Recently, Orka's research team has achieved significant accomplishments in the Second Clarity Enhancement Challenge (CEC2), winning the Bronze Prize, and the Silver Prize in the subsequent ICASSP SP Clarity Challenge.

The annual ICASSP challenge series, organized by the globally respected authority, the Institute of Electrical and Electronics Engineers (IEEE), aims to explore the frontier issues in the fields of acoustics, speech, and signal processing. The First Clarity Enhancement Challenge (CEC) was seen by technological powerhouses like Google as a crucial avenue for exploring and developing novel acoustic technologies, which prompted IEEE to launch CEC2 and the ICASSP SP Clarity Challenge. Both challenges focused on the application of the latest speech enhancement technology in the field of hearing aids, together promoting the development of speech enhancement functions of hearing aids, and attracting participation from excellent research teams globally.

Orka, a hearing aid company with co-founders from tech giants including Apple, Siemens, and Facebook, dazzled in these two challenges. With their outstanding research and development strength in the field of acoustic algorithms and signal electronics engineering, Orka team has performed well in both challenges, winning two awards for its strong research capabilities. In the CEC2 Challenge, Orka stood out from the 12 participating teams and won the bronze prize; later, in the ICASSP SP Clarity Challenge, Orka took second again, demonstrating their determination to innovate in related fields.

Figure 1 Results of CEC2

Figure 2 Results of ICASSP SP Clarity Challenge

It is worth noting that CEC2 and ICASSP SP Clarity Challenge were more complex than previous challenges, providing a test environment closer to real-life scenarios, simulating the head movement of listeners and various speech, music, and household noise sources. The competition systems and models underwent dual assessment through real-person hearing tests and objective indicators, measuring the effect of improving speech clarity without impairing sound quality.

Orka’s team optimized existing machine learning algorithms, with a particular focus on enhancing the amplification performance of hearing aids in real environments. They proposed a pure deep learning low latency hearing assistance system in real-time, including noise reduction module, beam forming module, and amplification module. In the CEC2 challenge, with an average HASPI score of 0.7673, Orka's model outperforms other state-of-the-art speech enhancement models like DCCRN [7] and SDD-Net [3]. Orka window also plays an important role in this experiment, simultaneously achieving high-frequency resolution and low reconstruction latency. In the ICASSP SP Clarity Challenge, Orka's HASPI and HASQI scores achieved an average of 0.398, securing second place—just shy of the first-place score of 0.403.

The success achieved by Orka in CEC2 and ICASSP SP Clarity Challenge signifies the company’s continuous commitment to research and development in the hearing aid industry. Along the path of actualizing frontier technology driven by the technical team, this accomplishment will contribute to further enhancing the performance and effect of hearing aid solutions. Orka will continue to work towards improving the quality and experience of hearing assistance devices, providing a better listening experience for users.