For pure intensity evaluation, a peak detector followed by some 'soft' low pass may come handy. But evaluation other than for intensity will not be possible on the filtered signal.

For all other purposes, a decent low pass filter may come handy if noise has to be reduced. Otherwise: no filter prior processing may prove the best filter. You will most likely require some filtering following signal processing - on the results' side.

Regards

If the recorded speech signal is degraded by Background noise, using wavelet tresholding would be more appropriate than low pass filters for vowels.