Apparatus, Method and Computer Program for Analyzing a Sensor Signal

Simple SummaryContent extracted from patent full text and abstract with AI.

The patent describes an apparatus, method, and computer program for energy-efficient classification of sensor signals, such as those from wearable devices monitoring physiological data (e.g., ECG for atrial fibrillation detection). The system uses a neural network whose weights are quantized to n-ary values (such as ternary: -1, 0, +1), allowing highly efficient computations with reduced memory and power requirements. Specialized hardware, including non-volatile multi-level memory cells (e.g., RRAM) and systolic array processor architectures, are used to further optimize energy consumption and device compactness, making continuous, long-term monitoring possible in wearable health devices.

Use CasesContent extracted from patent full text and abstract with AI.

• Wearable health monitors for continuous atrial fibrillation detection using ECG signals.
• Portable heart health screening devices for home or remote patient monitoring.
• Other types of wearable or implantable medical sensors (e.g., respiration, EEG, glucose monitoring) requiring efficient, real-time signal classification.
• Edge AI devices that process sensor data locally with limited battery or size, such as fitness trackers or medical alert systems.
• Industrial or environmental sensors where efficient, low-power signal classification is required.

BenefitsContent extracted from patent full text and abstract with AI.

• Significant reduction in energy consumption—up to 94.7% less than traditional processor architectures—enabling longer device operation without recharge.
• Smaller device size due to reduced battery requirement, supporting more comfortable and discreet wearables.
• Highly accurate and robust signal classification thanks to optimized neural network quantization and memory-aware training techniques.
• Support for non-volatile memory allows fast standby, instant-on operation, and retention of neural network parameters without constant power supply.
• Specialized architecture (systolic arrays, multi-level cell memory) makes the solution suitable for integration into various compact hardware platforms.
• Flexibility to adapt for different sensor types or medical applications via reprogrammable weights and devices.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

Embodiments according to the invention comprise an apparatus for analyzing, e.g. classifying, a sensor signal, e.g. a signal of a wearable sensor; e.g. an electrocardiogram signal or an ECG signal, e.g. for a detection of an atrial fibrillation on the basis of an electro-cardiogram. Furthermore, the apparatus is configured to input the sensor signal, or a preprocessed version of the sensor signal, or sensor data derived from the sensor signal, into a neural net (e.g. a neural network), coefficients, e.g. weights, of which are quantized to be n-ary weights, wherein, as an example, n is, preferably, a non-negative odd integer number which is larger than or equal to 3, e.g. ternary weights, e.g. to take three possible values or, e.g. 5-ary weights, which may, for example, take values of -2,- 1,0,+1,+2, or, e.g. 7-ary weights, in order to obtain an analysis result.