Methods and apparatuses for feature-driven machine-to-machine communications
摘要
說(shuō)明書
For example, a first application may process image data to count the number of trucks passing the crossroad, and a second application may process image data to detect identification (e.g., license plate number) of each passing vehicle. The second would require a higher image resolution than the first application. The typical simplest way to design a universal information representation and transmission scheme for transmission of information from the sensors is to adopt the most demanding requirement (e.g., highest image resolution required by the second application) across all sensors. However, the result is that there would be machine-perception redundancy for any application that has a lower requirement (e.g., the second application requires only a lower image resolution). In other words, the information that is transmitted to the second application has more information than required by the second application. The result is that the transmitted information does not make the most efficient use of channel capacity and power resources. However, there are difficulties to establishing an information representation and transmission scheme adapted for each application. These difficulties are similar to those encountered when considering time-related redundancy. For example, there may be little or no knowledge about the characteristics of the source information (i.e., a black-box information source) and there may be endless and unpredictable new updates to the observed subject. Moreover, the amount of machine-perception redundancy may be continuously changing. For example, an application may dynamically and automatically change its information needs. For instance, if vehicular traffic on the crossroad increases, the first application may require an increase in the resolution of the image data.
