News Mobile, IoT, Galaxy Watch, AI, Security

March 2024 Newsletter: Intelligence UX on Galaxy S24, Unsupervised ASR Personalization, and Other Latest News

Intelligence UX: Making Mobile AI Features Accessible and Intuitive for Everyone

Think of a world where you can talk to anyone without language barriers and illustrate your imagination with just a few touches. AI technology that once seemed a distant dream is now becoming part of our daily lives. Designers at Samsung Electronics have developed Intelligence UX (User Experience) to break down the towering wall between technology and daily life.

Meet the Intelligence UX designers and their thoughts on integrating AI technologies in users’ daily lives, and learn the story of Intelligence UX on Galaxy S24 series devices. Read the interviews at the Samsung Electronics Newsroom.

Learn more

Integrate Hub-Connected Devices with SmartThings Edge Device Builder

The SmartThings Edge Device Builder makes it easier than ever to integrate your Matter, Thread, Zigbee, and Z-Wave devices with SmartThings. You can quickly match your device with the optimal driver and profile and obtain Works with Samsung SmartThings certification. Learn more through the tutorial.

Learn more

Using Tizen Studio Features in Incompatible OS Environments

Tizen Studio provides a developer-friendly environment for debugging and testing for Tizen-based devices. However, if your computer’s operating system does not fully meet the requirements for Tizen Studio, some key features, such as Device Manager and Certificate Manager, may not work properly in the UI, causing difficulty with testing your applications on an emulator or target device.

This article describes workarounds for using Tizen Studio with an incompatible OS, and provides basic information about running and debugging applications through the command line interface. Learn more at the Samsung Developer Portal.

Learn more

Consistency-Based Unsupervised ASR Personalization

With voice assistants being widely used to control smartphones, automatic speech recognition (ASR) models play a critical role in recognizing the user's voice commands. On-device ASR models trained on speech data of a large population can underperform for users not sampled in the training, because of the domain shift between user data and the original training data. Most ASR personalization methods require labelled user data for supervised training, for which users must provide or revise transcripts.

This article introduces a paper accepted at IEEE ASRU 2023 that presents research on consistency-based unsupervised ASR personalization. Learn about unsupervised ASR personalization through pseudo-labelling at the Samsung Research blog.

Learn more

Multi-Modal Beam Selection: A Multi-Frequency Transfer Methodology

Wider application of millimeter wave and terahertz techniques have lifted constraints on spectrum resources in wireless communication networks. However, high-frequency electromagnetic waves in outdoor settings suffer severe path loss, resulting in cellular coverage reduction. Multiple-antenna and beamforming techniques are important in enhancing throughput.

Existing analog beamforming methods guarantee beam selection accuracy at the expenses of time and transmission cost, particularly in massive multiple-input and multiple-output (MIMO) scenarios. This article presents a multi-modal beam selection method and a frequency-transfer framework for training models that improve on existing methods through assistance in optimal beam selection and extraction of environment and path loss information. Learn about them at the Samsung Research blog.

Learn more

KFLAT: Selective Kernel Memory Serialization for Security and Debugging

The Linux operating system is very complex software and developing for it requires advanced testing and debugging capabilities. Memory dumps, such as a core dump in GDB or a memory image fed into a VM, are the main sources of data for these tasks. However, common memory dump tools lack granularity. They may dump an entire process or system memory without discriminating which source code structures it represents.

The Mobile Security Group at Samsung R&D Institute Poland has developed KFLAT, a tool that creates a fine-grained copy of kernel memory for selected variables and structures. This copy can be used to recreate the kernel memory layout in the user space process on another machine. Find out more about KFLAT, how it works, and how to use the tool at the Samsung Research blog.

Learn more

March 2024 Newsletter: Intelligence UX on Galaxy S24, Unsupervised ASR Personalization, and Other Latest News

Intelligence UX: Making Mobile AI Features Accessible and Intuitive for Everyone

Integrate Hub-Connected Devices with SmartThings Edge Device Builder

Using Tizen Studio Features in Incompatible OS Environments

Consistency-Based Unsupervised ASR Personalization

Multi-Modal Beam Selection: A Multi-Frequency Transfer Methodology

KFLAT: Selective Kernel Memory Serialization for Security and Debugging

April 2024 Newsletter: Integrating Services to Samsung Wallet, and Other Latest News

Galaxy A55 5G and A35 5G Emulator Skins are Available for Download

February 2024 Newsletter: Galaxy AI-Assisted Tools that Maximize Workplace Productivity and Other Latest News