A problem of signal estimation (or estimation of some parameters of the signal) based on measurement results obtained by means of distributed sensor network is considered. The system formed by spaced sensors called nodes (or motes) obtains information about a stochastic process. It is supposed that there is a single source of the information and the distortions are linear.
Every node gets incomplete and distorted knowledge about the observed process and operates on the low sampling rate. Combined processing of the information obtained from the different sensors makes possible to perform the integrated estimation which is usually much more accurate.
The application of the information performance criterion provides high robustness in the presence of the noise and the measurement uncertainty and, theoretically, gives the most complete information, while the measurements have severe constraints.
Due to the multirate signal processing it is possible to decrease requirements on sensor’s sampling frequencies, to reduce the channel loading and the current consumption of sensors. Furthermore, the computational load falls in real-time applications.
Keywords: multirate signal processing, sensor, node, mote, information, stochastic process
The new way of the organization of multistage time-frequency signal localization subsystems is considered. The investigations are constrained by finite impulse response class of digital circuits.
The example illustrated the principle of agreement between pyramidal structure, formed of multistage including of any simple analysis/synthesis subsystem, and channel characteristics. This principle makes possible more precise localization of the input signal in separate frequency bands. The adaptive structure formation of the time/frequency transformation allows increasing of parameters characterizing the effectiveness of the subband adaptive filters (SAF).
Application of the method of flexible time/frequency localization structure formation in the adaptive filtering problems allows significantly decreasing of processor loading by real-time realization of SAF.
The other important advantage of the flexible SAF formation based on the least-mean-square algorithm lies in reducing of the adaptation period for the adaptive system.
Keywords: adaptation, filter, multirate, signal, time/frequency
The new way of multirate adaptation is considered. The main idea is the filter ability to change the order of the filter automaticaly by the appropriate algorithm.
The power on the filter output consists of two components. The first one is caused by the impulse response truncation, and the other is caused by weights fluctuations in the adaptive processing. More weights contribute more noise in adaptive signal processing. The result of adaptive procedure is the choice of the best filter orders in every channel of the multi-channel structure.
The main interest of this investigation is in the adaptation of the multirate adaptive filter with the flexible structural self-organization.
Due to the use of dynamical order adaptation procedure in the subbands, it is possible to reduce the total amount of the computational operations without quality reducing. Beside it, decreasing of the filter orders in the separate channel promotes considerable increase of the signal-to-noise ratio in the output signal. These statements are approved by experiments with the equalization problem.
Different results of the computer modeling are supplied with proper comments.
Keywords: adaptation, filter, multirate, signal, dynamical, order