Methods and apparatuses for feature-driven machine-to-machine communications
摘要
說(shuō)明書
At 802, the BS 120 receives raw information that has been collected and transmitted by each sensor 110 about the observed subject 105. If the training takes place in another component of the core network 130 or outside of the core network 130 (e.g., at a remote data center 160), the BS 120 may further transmit the raw information to the appropriate entity. For simplicity, the present discussion will refer to the example where the training is performed at the BS 120. However, it should be understood that this is not intended to be limiting, and steps of the method 800 may be performed elsewhere in the network (e.g., at another component of the core network 130 or other network entity).
The raw information may be stored as training samples, for example in a local memory or remote database (e.g., at the remote data center 160) accessible by the BS 120. Training of the encoder and decoder DNNs may be done for one type of sensor at a time (e.g., sensors gathering visual information, or sensors gathering audio information), in which case the raw information may be collected only from one type of sensor 110 connected with the BS 120. For example, the BS 120 may assign a sensor type to each connected sensor 110 (or each sensor 110 may declare its own type) and may request raw information from one type of sensor at a time. Alternatively, the BS 120 may receive raw information from all sensors 110 regardless of type, and the BS 120 may organize the raw information into separate sets of training samples according to sensor type.
