Linear predictive coders are a kind of significant vocoders. The speech coders using three linear predictive vocoding algorithms are described, i.e. CELP (code excited linear prediction), LD-CELP (low delay-CELP) and MELP (mixed-excitation linear predictive) working at 4.8 kb/s, 16 kb/s and 2.4 kb/s, respectively. The C programs for the three vocoders are complied and executed on the Linux platform and the results are subjectively tested with taking MOS as criterion. Waveform analysis mainly uses Praat and Adobe Audition software. The results show that the performances of MELP and CELP are of comparable quality, while the performance of LD-CELP is much higher, albeit at the expense of higher bit rates.