The invention provides a speech synthesis device, a speech synthesis method, a bone conduction helmet and a hearing aid, and belongs to the technical field of communication equipment. The problem that the accuracy of speech signals of the existing speech device is low can be solved. The speech synthesis device of the invention comprises at least one sink node and multiple detection nodes, wherein the detection nodes are used for detecting external speech signals in real time and transmitting the external speech signals to the at least one sink node, and the at least one sink node is used for screening the collected speech signals and synthesizing the speech signals sent by trusted detection nodes. Thus, the bit error rate of speech signals can be reduced greatly, and the accuracy of speech signals can be improved.