Trout GmbH
Parkstr. 28
34119 Kassel (Deutschland)
Tel: +49 561 810497-0
Fax: +49 561 810497-15

Kontinuierliche Bestimmung des Herzschlags auf Basis von Radardaten unter Verwendung von Deep Learning Methoden

news imagedownload
With years of experience in biometric data processing, TROUT has extensive knowhow in incorporating artificial intelligence (AI) and machine learning in their innovative products and solutions. Two notable examples are VitaB where machine learning is used to acquire vitality parameters and to classify cognitive condition; and NivaB where non-invasive determination of blood glucose is performed by impedance spectroscopy and the application of neuronal networks.
At present, TROUT is using a deep learning approach to predict single heartbeats (R peaks) in time series data from a low energy radar measurement reflected by the human chest. The goal is to estimate heart rate variability (HRV), a vital parameter in health care.
An ECG signal is used to label single heartbeats in the radar data. The prediction utilizes a long short-term memory (LSTM) cell-based neural network model. The LSTM cells realize a recurrent feedback mechanism, which allows the evaluation of live data in real time. LSTM implements the groundbreaking approach of storing past experiences in a short-term memory that lasts for a long time.
As we continue to optimize our LSTM implementation, we will extend the method to forecast additional features of the ECG trace and determine further cardiac parameters such as Q and S peaks.

zurück zu den Ausstellerdetails