A streaming end-to-end speech recognition method and apparatus, and an electronic device. The method comprises: performing speech acoustic feature extraction and encoding on a received speech stream in a unit of frame (S301); dividing an encoded frame into blocks, and predicting the number of activation points that are comprised in a same block and need to be encoded and output (S302); and determining, according to the prediction result, locations of activation points that need to be decoded and output, so that a decoder can perform decoding at the locations of the activation points and output a recognition result (S303). The method can improve the robustness of a streaming end-to-end speech recognition system in terms of noise, thereby improving the system performance and accuracy.