Some medical products are reaping the benefits of DeepSeek’s advancements. One example? Consider hearing aids.
Although hearing aids have been around for decades, restoration of auditory capabilities in certain environments—like bustling marketplaces or noisy parks—remains a challenge.
As products that heavily rely on algorithms and circuitry, hearing aids have incorporated machine learning for many years. But now, DeepSeek and other open-source large models present a new opportunity for manufacturers to overcome longstanding challenges in hearing aid development: providing “a chance to leverage the unique technical insights of world-class large models at low cost,” said Wang Fa, president of United Imaging Healthcare’s hearing aid division.
Globally, the hearing aid market has long been dominated by five established players, including Switzerland’s Sonova, Denmark’s Demant, and Starkey from the US, and more. They collectively account for over 90% of total sales. In recent years, a wave of tech startups and healthcare companies has been entering the space, but they continue to grapple with challenges like tech-related limitations and chip supply.
So, could the integration of large models and the enhancement of artificial intelligence capabilities in hearing aids offer newcomers a chance to break through?
Using AI to restore directional hearing
Compared to other consumer-grade medical products, the overall penetration rate of hearing aids—and the actual benefits they bring to users—remains relatively low. A previous survey conducted by Peking University’s Institute of Population Research showed that 90% of elderly people with hearing loss, despite being aware of their condition, were unwilling to undergo hearing tests. Among those who did opt to use hearing aids, only 6.5% were satisfied with the results.
Why are people reluctant to wear hearing aids? And why do those who buy them stop using them?
Liang You (pseudonym), a veteran in the hearing industry with over two decades of experience at a top-tier hospital in Beijing, attributes the problem to several factors. Beyond the historical neglect of otology in hospitals and limited public awareness, the underperformance of traditional hearing aids remains a key issue.
“Objectively, today’s products are much better than those from the 1990s, but many hearing aids still struggle to identify sounds in noisy environments. This leaves elderly users feeling that the device is useless, leading them to abandon it after a few attempts,” Liang You said.
Those who look to buy hearing aids for their elderly relatives echoed similar sentiments, saying that their loved ones often react with skepticism, believing that the devices don’t work or complaining about excessive background noise after a few tries, causing them to give up.
At its core, this perception of ineffectiveness stems from a classic problem in speech recognition: the “cocktail party effect.” When multiple sound sources exist in a given environment, how can a hearing aid distinguish the voice of a specific individual?
According to Wang, for decades, hearing aids relied on digital signal processing (DSP) algorithms, which operate on numerous assumptions while addressing noise reduction. For example, these algorithms often assume that human voices should be preserved while background noise should be filtered out. However, in real-life noisy environments, background sounds may also include voices, and traditional algorithms inadvertently retain these sounds. As a result, when multiple competing voices reach the user’s ears, the ability to discern speech is compromised.
“The industry has long been caught between the difficult trade-off of noise elimination and preserving speech detail,” Wang said. “This is where the world’s leading hearing aid manufacturers are focusing their technical efforts.”
In the past two years, both the “big five” manufacturers and emerging Chinese brands have increasingly recognized that the key to solving this challenge lies in introducing deep neural networks (DNNs), which is a machine learning technology that mimics the structure and functionality of human neurons. DNNs enable hearing aids to move beyond traditional assumptions, continuously learning and fine-tuning based on the vast amounts of real-world audio they encounter daily, thereby improving speech recognition in different environments.
This advancement promises to deliver the greatest benefit to hearing-impaired individuals.
“While product size, power consumption, and battery life are also areas of technical competition, none are as directly impactful as solving the problem of directional listening in noisy environments,” Wang said.
Moreover, large models represented by DeepSeek can be deployed to adapt to users’ environments. This enables hearing aids to perform adaptive learning, self-fitting, and intelligent fitting based on the user’s surroundings.
The improved product experience brought by these AI algorithms has become a major focus for leading hearing aid manufacturers.
In late February, Phonak, a subsidiary of Sonova, officially launched an AI-powered hearing aid product in China that integrates DNN processing algorithms, achieving a noise reduction effect of ten decibels in complex environments. Compared to the industry average, where noise reduction levels typically range between four to six decibels, this advancement is significant. Around the same time, Starkey also announced updates to its flagship “Genesis AI” hearing aids, incorporating DNN processors that simulate the brain’s central auditory system to automatically detect surrounding environments.
On the domestic front, United Imaging’s uOrigin has adopted its self-developed heterogeneous six-core chip and edge-side neural network algorithm, capable of processing up to 750 sound signals per second. It boasts smart noise reduction and directional sound pickup capabilities. Additionally, emerging hearing aid brands that have attracted considerable capital, such as Elehear, are also making DNN technology a core aspect of product development.
Interestingly, it’s not just traditional medical companies venturing into the hearing aid space, as tech giants are also making moves. Apple introduced hearing testing features in the AirPods Pro 2, while Tencent and iFlytek are actively exploring the mid-range hearing aid market with products priced around RMB 1,000 (USD 140).
“In the realm of wearable health products, tech companies have already saturated the market with entry-level functions like heart rate monitoring, blood sugar tracking, and sleep quality assessment. Given the technological similarities between hearing aids and traditional true wireless Bluetooth earbuds, it’s only natural for hardware giants to enter this space,” said a former engineer from a leading international tech company.
Consumer-grade hearing aids: A chance for newcomers?
As AI levels the playing field by narrowing the technology gap, could the new generation of hearing aid companies get closer to challenging the dominance of the “big five?”
One potential advantage lies in the increased public awareness of hearing aid products after years of market education by industry pioneers. For instance, some e-commerce insiders have observed a decline in the number of consumers opting for low-cost sound amplifiers and a growing preference for professionally fitted hearing aids.
Meanwhile, the emergence of new manufacturers provides consumers with more choices, including the popularity of over-the-counter (OTC) hearing aids.
To understand this shift, it’s essential to consider the revenue model of hearing aids. The primary distinction between medical- and consumer-grade products is whether professional fitting is required. Medical-grade hearing aids are typically purchased from brick-and-mortar stores, where audiologists fine-tune the device to match the user’s individual needs. This process can take three to six months. In contrast, consumer-grade hearing aids can be bought and used directly online.
Price-wise, industry insiders generally consider RMB 4,000 (USD 560) as the dividing line. Medical-grade hearing aids from the “big five” typically sell for RMB 5,000–40,000 (USD 700–5,600) per unit (with potential discounts), while top-tier Chinese brands sell for RMB 5,000–20,000 (USD 700–2,800). Meanwhile, consumer-grade hearing aids are priced RMB 1,500–4,000 (USD 210–560).
The significant price gap is partly due to the cost of professional services. “Compared to consumer-grade products, medical-grade hearing aids incorporate more advanced algorithms and applications, making the hardware cost slightly higher. But to be honest, the cost difference alone doesn’t justify a tenfold increase in price. Much of the price difference is attributed to the value added by professional services,” the former engineer said.
Although there have been voices within the industry suggesting that products and services should be charged separately, this is difficult for manufacturers to implement, as there is limited acceptance of medical service fees in China.
Some manufacturers believe that by eliminating this cost, consumer-grade hearing aids can potentially capture a share of the market through a low-cost strategy veering close to the RMB 1,000 range.
For example, iFlytek mentioned in its prospectus that revenue from its smart hardware segment reached approximately RMB 6.2 million (USD 868,000) in the first half of 2022 and surged to RMB 29 million (USD 4.1 million) in the first half of 2023, driven in part by the launch of its hearing aid product in May 2022. As of early this year, iFlytek’s registered smart hearing aid users have reportedly exceeded 110,000.
However, opting for the consumer-grade route doesn’t guarantee success. In Liang You’s view, the ability to purchase and fit hearing aids independently does not absolve manufacturers of the need to provide professional support.
“Since these are still hearing aids, companies must offer follow-up services to ensure that the product does not harm the user’s hearing during use. Moreover, when users encounter issues, manufacturers should be ready to provide remote assistance,” Liang You said.
Wang echoed this sentiment, stating that while consumer-grade hearing aids will “undoubtedly become an integral part of the market,” current products still fall short of fully replacing professional audiologists. “They can handle some basic functions like hearing assessment, fitting, and usage, but they are still a long way from acting as virtual audiologists capable of providing comprehensive services.”
At the end of the day, while hearing aids carry a “medical” label, purchasing decisions are heavily influenced by individual consumer preferences and needs. In the short term, the market structure dominated by the “big five” is unlikely to be disrupted. However, in the longer term, factors such as product pricing, the development team’s understanding of the language environment, and the quality of after-sales service will likely become key differentiators.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Hu Xiangyun for 36Kr.