The invention discloses a rapid recognition method for incomplete voices of old people, which comprises the following steps: firstly, training a DNN-HMM voice recognition model, then collecting
voice data of the old people, sequentially carrying out preprocessing,
noise reduction enhancement, stop word removal and keyword expansion on the data, constructing a voice
database of the old people, and carrying out word segmentation and word vector calculation on statements in the
database; and finally, converting the incomplete speech of the old people by using a DNN-HMM
speech recognition model, carrying out word segmentation and word vector calculation, then matching with statements in a
database, and outputting a statement with the highest similarity, thereby completing rapid recognition of the incomplete speech. According to the technical scheme, only a
microphone or other voice acquisition equipment is needed to acquire the voice of the elderly, electronic detection equipment such as an
electrode patch is not needed, the influence on the life of the elderly is small, the recognition speed is high, the accuracy rate is high, and the method has very high popularization value.