This paper presents an implementation on FPGA of a one dimension discrete wavelet transform (1DWT) for simulation of a medical application dealing with hearing loss. The Digital DWT (DDWT) can be viewed as a multi-resolution decomposition of a signal which means that it decomposes the signals into its principal components in different frequency bands (octaves bands). This configuration is matching model to our designed artificial cochlear membrane (ACM).
It is proposed a new design approach of an architecture based on parallel FIR filters by considering the implementation of a three level DWT. The proposed architecture offers 16 bits on input and output data, yielding to high performance precision.
It is constituted on register banks, a control unit and FIR filters.
The filters are of different lengths and with coefficients derived from FIR filter coefficients using Matlab / Simulink. The designed system can compute DWT at a typical frequency clock over 100 Mhz which easily reach a real time processing for speech enhancement.
The system is evaluated using VHDL implementation over Virtex 6 (XC6SLX45T) of Xillinx which includes 58 (DSP48A1) slices. The results are compared to that obtained by Matlab simulation.