Matched Filters in Nerve Conduction Velocity Estimation

The evoked response signal-to-noise ratio in peripheral sensory nerves is of the order of one or less. To reduce noise induced errors in the nerve conduction velocity measurement, signal averaging is employed. The number of responses required depends upon the signal-to-noise ratio and the acceptable error. It is proposed to use maximum likelihood estimators to improve the signal-to-noise ratio and hence, reduce the number of responses required. The matched filter and generalized cross correlator are studied and noise performance equations obtained. It is shown that the matched filter offers a 6.8 dB improvement over the conventional 5 kHz band-limiting filter. It is further shown that the matched filter as a prefilter in the generalized cross correlator gives the optimum correlator configuration for the low signal-to-noise ratio case. These results are verified with experiments.