This paper investigates the wavelet packet based front-ends for the connected mandarin digit recognition task. Firstly an ERBlike wavelet packet basis is proposed. Then two kinds of wavelets are selected for comparison. One is the Vaidyanathan wavelet, which has good frequency selectivity but big shift variance. The other is the reverse biorthogonal spline wavelet with excellent shift invariant property. Thirdly, the Teager-Kaiser energy operator (TEO) based subband cepstral (TC) feature parameters are extracted from the wavelet packet derived multi-frequency channels. The recognition results of the new front-ends are tested and compared with the popular MFCC parameter on the 8K 16-bit speaker-independent mandarin connected digit corpora. Apart from clean data condition, the performances of the new front-ends are further compared in various noisy conditions.