Highlights 2025 – Annual Report

Over the past decade, our researchers have combined their expertise in digital signal processing with advanced AI models. This has led to two new technologies that have been in use since 2025.

One of these solutions is Fraunhofer NESC, which marks a new generation of AI-based voice codecs. Conventional voice codecs used in 4G and 5G services typically require more than 10 kbit/s to achieve high audio quality and speech intelligibility. By leveraging AI, NESC reduces the required bitrate for excellent speech quality to just 1-3 kbit/s. This enables network operators to provide reliable voice services over new distribution paths, including satellite links. In 2025, we demonstrated the first voice services over non-terrestrial networks (NTN), working with Bullitt and Skylo amongst others.

Another breakthrough is upHear Personalized Voice Isolation, an AI-enhanced technology that learns a speaker’s voice and generates a digital fingerprint of it within seconds. This allows the system to enhance the target voice while suppressing others. The approach addresses challenges that conventional signal processing methods have been unable to overcome for decades. The benefits are particularly evident in online meetings held in environments with multiple simultaneous conversations, where microphones capture several speakers at once. upHear Personalized Voice Isolation isolates the desired speaker’s voice and at the same time reduces other voices.

Be it smartphones, medical technology, or AI – microelectronics is the foundation of it all. It has always been at the core of our Smart Sensing and Electronics division, which also coordinates the Bavarian Chip Design Center. In 2025, the Center celebrated the successful launch of its trainee program.

For six months, seven talented young professionals immersed themselves fully in chip design. Based on their interests, they trained in one of three focus areas: Analog IC Design, Test and Validation, or System Engineering. Working in interdisciplinary teams, they contributed to real projects – designing systems, simulating circuits, and inspecting chips down to the smallest detail.

The program prepares specialists for a key enabling technology, as microelectronics continues to shape the digital future. Our trainees not only deepen their technical expertise but also gain firsthand insight into how research and industry work hand in hand to bring innovations into application quickly. Companies benefit directly from practically trained experts who can tackle complex challenges independently.

After a successful launch, the trainee program will enter its next round in April 2026, now expanded to include a focus on Digital IC Design. For the first time, companies can now nominate their own employees, who will likewise be prepared to help shape the future of microelectronics and drive chip innovation forward. In doing so, we are strengthening the next generation of skilled professionals and bolstering Germany’s competitiveness in a data- and chip-driven world.

In Germany, more than 10,000 fires occur each year in waste sorting facilities – often because improperly disposed lithium-ion batteries end up in the recycling bin. According to the industry association BDE, around 80 percent of these fires can be traced back to this issue, causing damages of roughly one billion euros annually. We therefore intervene earlier by identifying critical batteries and removing them from the material stream before they ignite.

Within the K3I Cycling project, we developed a sensor-based sorting system. An X-ray unit – similar to an airport baggage scanner – scans the waste on the conveyor belt. Using AI-based, real-time image analysis, the system detects both visible and hidden batteries. Precision air nozzles then separate the identified items from the rest of the material. This increases safety for employees and facilities while also improving battery recycling.

The wide variety of battery formats – from e-bike packs to button cells – as well as the millisecond ‑accurate timing required between image analysis and air nozzle‑ triggering present significant challenges. Thanks to training with real-world‑ data, detection rates and system robustness continue to improve.

Since June 2025, we have been testing the system under real operating conditions together with LOBBE RSW GmbH. Funded by the Federal Ministry of Research, Technology, and Space and embedded in the AI Hub for Plastic Packaging, we are working on safe, resource efficient recycling loops. As device numbers continue to grow, early detection and separation of hazardous objects is becoming the standard – today and in the years to come.

Rising costs, intense competitive pressure, and rapid technological change are posing major challenges for small and medium-sized enterprises (SMEs). While new AI-based technologies offer opportunities for production, SMEs often perceive them as complex. Yet adapting to these developments is essential, especially for SMEs, to remain efficient and competitive in the long term.

At our Dresden site, we have built-up a demo production environment where AI applications can be tested under realistic conditions. Companies can gain their first hands-on experience or explore and further develop specific application scenarios. This provides a central hub that makes the use of AI more accessible and enables rapid knowledge transfer.

The existing infrastructure is used to address individual challenges in a safe environment and with expert guidance. In the first stage of expansion, SMEs can work through real production processes at five stations – ranging from adaptive robotics, and wireless communication to smart sensor technology. The potential of edge components and a digital twin is also showcased.

The demo production environment is designed as a flexible platform that can be tailored to the specific needs of companies. Through open exchange and hands-on technologies, we support businesses in lowering innovation barriers and actively advancing digital transformation.

This measure is co-funded by tax revenue based on the budget approved by the Saxon State Parliament.

An independent, reliable global positioning and timing solution is more important today than ever before. In 2025, Galileo – the European Union’s satellite navigation system and the world’s only fully civilian-operated service – celebrated its 30th anniversary.

In its early years, satellite navigation was limited to professional and costly applications. Twenty years ago, in-car navigation devices began to reach the mass market. With the introduction of smartphones, satellite navigation became an everyday utility, enabling everything from high-precision time synchronization to telecommunications and energy systems, as well as financial markets and banking.

We have been working intensively on Galileo – and with it – for 25 years. Our applications range from developing high-precision navigation receivers, for example, for forestry machinery operating in wooded terrain, to ship-control systems and receiver technology for research rockets and satellites. The EU’s Galileo program also launched a research funding initiative for satellite navigation technologies in Europe, enabling our researchers to put new ideas and concepts into practice.

By developing a specialized monitoring station for the EU’s Galileo Public Regulated Service (PRS) – an encrypted, highly protected service for governmental and security-critical applications – we are making a significant contribution to enhancing the security and resilience of our critical infrastructure.

The future of industry is digital and connected – not only within individual companies, but across entire value chains. In many areas, however, the reality still looks different: isolated systems and data silos hinder efficient processes and the development of new products and services.

In drive technology in particular, the idea of cross-company data exchange has yet to take hold. Reasons include concerns about maintaining secure control over proprietary data and a lack of concrete ideas for practical use cases.

In the Antrieb 4.0 project, we are therefore working with research partners and drive manufacturers to develop a solution that enables interoperable, cross-manufacturer data exchange without requiring companies to give up data sovereignty. To achieve this, we rely on Semantic Web concepts and decentralized Solid technology: data is described in a standardized way, stored in a dedicated data vault, and can be shared selectively. This makes it possible to optimize production systems across manufacturers and operators, build digital twins with real-time data, or implement energy-efficient drive systems – use cases we envision within a service platform for electric drives. With Antrieb 4.0, an easy-to-implement, streamlined solution for digital, cross-company services is taking shape – built on the same principles as more complex dataspace initiatives such as the GaiaX Federation Services. This makes it easier for companies to enter an increasingly connected data ecosystem, while leaving ample room for future expansion.

When people channel all their energy into an explosive burst of power at the start of a sprint, a firework of electrical impulses ignites in the brain. These pulses race through networks of neurons, activating precisely the right actions. This same principle underlies the next generation of energy-efficient artificial intelligence: Spiking Neural Networks (SNNs). Until now, however, there has been very little specialized, brain-inspired hardware capable of harnessing this neuron-driven rapid-response capability for communication systems, industrial applications, or sensor-based devices. That is now beginning to change.

With the development of the SNN processor SENNA, we have succeeded in capturing the electrical replica of brain-like neurons on a microchip. The current SENNA chip contains 1,024 such artificial neurons and analyzes signal streams within nanoseconds. Building on this capability, it can enhance data transmission, regulate and control small electric motors, or respond to events inferred from sensor data at lightning speed. To integrate SENNA’s pulsating precision into devices and systems, we support our partners from the initial feasibility study through to series production. SENNA’s latest task in the EU project SpikeHERO: eliminating signal distortions that occur at very high transmission rates, thereby making fiberoptic networks even faster.