An SoC Architecture for Real-Time Noise Cancellation System Using Variable Speech PDF Method

Abstract

This paper presents the architecture and implementation of system-on-chip (SoC) for realtime noise cancellation system which exploits variable speech probability density function (PDF) and maximum a posteriori (MAP) estimation rule as noise cancelling algorithm. The hardware software co-design approach is employed to achieve real-time performance while considering ease of implementation and design flexibility. The software module utilizes LEON SPARC-v8 and FPU co-prosessor as processing unit. The AMBA based Hanning Filter and FFT/IFFT are utilized as processing accelerator modules to increase system performance. The FFT/IFFT module employs custom Radix-2^2 Single Delay Feedback (R2^2SDF). In order to deliver high data transfer rate between buffer and hardware accelerators, the DMA controller is incorporated. The overall system implementation utilizes 18,500 logic elements and consumes 21.87 kB of memory. The system takes only 0.69 ms latency which is appropriate for real-time application. An FPGA Altera DE2-70 is used for prototyping with both algorithms and the noise cancellation function have been verified.