Abstract: An audio monitoring system provides assistance from an external caregiver to a monitored person in need as efficiently as possible, and limits false alarms. To achieve this, audio signals in a monitored environment are continuously detected by a monitoring system and are explicitly not stored if there is no emergency situation. The incoming audio signal is analysed and classified by noise and speech recognition methods. If an emergency situation is identified, an audio recording is started, and a local alarm is initially sounded which the monitored person can acknowledge within a defined period of time. A remote caregiver is only contacted if no acknowledgement occurs. This method limits the number of false alarms. The audio recording is made available to the remote caregiver, who can therefore assess the emergency situation and development of the situation. Furthermore, an audio connection to the monitored person is established.