HPM 16.2 VLSI Implementation of 3-D Sound Generator

Summary In this paper we propose the VLSI architecture for the 3-D sound generator. Fig. 1 shows the editable 3-D sound system. First, the software simulation is developed. There are totally 7 10 head related transfer functions (HRTFs) measured from the MIT Media Lab[ I]. We use clustering method to sample the 710 sets of HRTF. The size of storage to store HRTFs is hence largely reduced by using only 157 sets of HRTFs. Because HRTF sets are discrete in space, audible clicks will occur. Crossfading algorithm is employed to eliminate audible clicks. Crossfading is a skill of interpolation in time domain which can be used to eliminate the impact of the clustering method. Each HRTF filter length is 128 points. The full transformation from mono sound to 3-D sound has heavy computational load due to the convolution operations between the mono input vectors and the HRTF filters. Hence to design a low cost single chip for generating the 3-D sound is becoming important. The 3-0 sound generation chip can be used in variety of fields in audio, such as commercial music recording, video game enhancement, virtual reality systems[2]. This paper presents the high performance VLSI architecture for the 3-D sound chip. A single 3-D sound chip design drastically reduces the cost and the size of inany audio systems. The proposed 3-D sound VLSI