RF fingerprint signal processing device and RF fingerprint signal processing method

1. An RF fingerprint signal processing device configured for executing a machine learning algorithm on a plurality of input signals, wherein the RF fingerprint signal processing device comprises:

a receiver-feature determination circuit configured to compute on the plurality of input signals in a neural network, wherein the input signals are radio frequency (RF) fingerprint signals;

a classifying determination circuit coupled with the receiver-feature determination circuit, wherein the classifying determination circuit is configured to send feedback information of a receiver-feature component to the receiver-feature determination circuit;

a transmitter-feature determination circuit coupled with the classifying determination circuit, wherein the transmitter-feature determination circuit is configured to increase only a transmitter-feature weight of the neural network, and the transmitter-feature weight is associated with a transmitter-feature component, the transmitter-feature determination circuit applies the transmitter-feature weight which is increased to compute an output value of the neural network; and

a cluster analysis circuit coupled with the classifying determination circuit,

wherein the receiver-feature determination circuit decreases only a receiver-feature weight of the neural network, the receiver-feature weight is associated with the receiver-feature component, and the receiver-feature weight which is decreased is applied for computing the output value of the neural network,

wherein when the classifying determination circuit outputs a training result of the neural network to the cluster analysis circuit, the cluster analysis circuit is configured to recognize a receiver source and a transmitter source of an unknown RF signal according to the neural network which has trained neuron weights.

2. The RF fingerprint signal processing device of claim 1 , wherein if the classifying determination circuit applies the output value to classify a feature signal of a first transmitter into a feature signal of a second transmitter, the classifying determination circuit analyzes the receiver-feature component of the output value to send feedback information of the receiver-feature component to the receiver-feature determination circuit.

3. The RF fingerprint signal processing device of claim 1 , wherein if the classifying determination circuit applies the output value to classify a feature signal of a first transmitter into a feature signal of a second transmitter, the classifying determination circuit analyzes the transmitter-feature component of the output value to send feedback information of the transmitter-feature component to the receiver-feature determination circuit.

4. The RF fingerprint signal processing device of claim 1 ,

wherein when accuracy of a determination of the transmitter source of the plurality of input signals made by the classifying determination circuit is large or equal to a threshold, the cluster analysis circuit outputs the training result of the neural network to the cluster analysis circuit.

5. An RF fingerprint signal processing method configured for executing a machine learning algorithm on a plurality of input signals, wherein the RF fingerprint signal processing method comprises:

computing on the plurality of input signals in a neural network, wherein the plurality of input signals are radio frequency (RF) fingerprint signals;

sending feedback information of a receiver-feature component;

increasing only a transmitter-feature weight of the neural network, wherein the transmitter-feature weight is associated with a transmitter-feature component;

decreasing only a receiver-feature weight of the neural network, wherein the receiver-feature weight is associated with the receiver-feature component;

computing an output value of the neural network by the receiver-feature weight which is decreased and the transmitter-feature weight which is increased; and

outputting a training result of the neural network, and recognizing a receiver source and a transmitter source of an unknown RF signal according to the neural network which has trained neuron weights.

6. The RF fingerprint signal processing method of claim 5 , further comprising:

analyzing the receiver-feature component of the output value if a feature signal of a first transmitter is classified into a feature signal of a second transmitter by the output value; and

sending feedback information on the receiver-feature component.

7. The RF fingerprint signal processing method of claim 5 , further comprising:

analyzing the transmitter-feature component of the output value when classifying a feature signal of a first transmitter into a feature signal of a second transmitter by the output value; and

sending feedback information of the transmitter-feature component.

8. The RF fingerprint signal processing method of claim 5 , further comprising:

outputting the training result of the neural network when accuracy of a determination of the transmitter source of the plurality of input signals is large or equal to a threshold.