VHDL implementation of Hopfield neural network acting as detector for buried Ferro-metallic materials

In this work a VHDL-based Hopfield neural network have been designed and applied to the detection of the buried ferro-metallic materials. The energy function of the proposed network is designed to optimize the magnetic moment of the dipole source representing the magnetic object at regular locations. For each location, the Hopfield neural network reaches its stable energy state, where the object position can be estimated from the output of the network at this state. The obtained energy function of the network includes too much iteration, mathematical functions such as sinusoidal, and both the division and square root operations. Implementing this energy function with VHDL generates chip with large size and long processing time. To optimize the size and speed of the chip, reduced list of the network weights is used. Also, the Taylor series of the sinusoidal function is modified to limit its exponent to 2. Moreover, the square root and division operations are implemented with successive approximation algorithm, which can successfully compute the value of these functions in a shorter time and smaller chip size. After applying these modifications, 31% of the chip size is saved and 20% of the processing time is reduced. It is also proved that the proposed network can locate the position of the buried objects quite accurately.