This paper presents a structural design for hardware implementation of low-bit-rate speech coding. The processing algorithm emerges from the traditional pitch-excited linear prediction vocoder with incorporation of glottal-source properties. To make the coding algorithm realizable, constraints are imposed on the complexity of computational modules as well as the need of arithmetic operations. Nonetheless, the simplified algorithm is still capable of carrying out 1.6Kbps speech coding with acceptable quality.We have gone through algorithm modification, fixed-point analysis, hardware design, and circuit synthesis by making use of various design tools. The hardware design, described by using Verilog-HDL, is compiled and verified on an FPGA-based DSP board. Our results confirm that the proposed architecture works quite well. As this architecture needs not consume many hardware resources, it is suitable for single chip implementation.