The implementation of an adaptive Volterra filter to compensate for speaker non-linearity using a pipelined recurrent neural network based architecture is demonstrated. The proposed architecture consists of two stages: nonlinear stage performing a nonlinear second order Volterra (SOV) mapping from the input space to an intermediate space and a linear combiner performing a linear mapping from the intermediate space to the output space. The filter design is tested by implementing it on a dataplane processing unit configured for audio processing. The implemented algorithm is adjusted for Xtensa Processor and uses HiFi-2 DSP standard. The collected data confirms smaller error of the output with the speaker and also a very low settling time.