Speech recognition model training method, speech recognition method, device and equipment

The invention provides a speech recognition model training method, a speech recognition method, a speech recognition device and speech recognition equipment. The training method comprises the following steps: acquiring a target type of a processor in electronic equipment; determining a target initial acoustic model corresponding to the target type, wherein the model structure of the target initial acoustic model is matched with the target type; training the target initial acoustic model based on a target corpus to obtain an acoustic model; and determining a speech recognition model based on the acoustic model. The adaptability of the speech recognition model can be improved. 本发明提供一种语音识别模型的训练方法、语音识别方法、装置和设备，该训练方法包括：获取电子设备中处理器的目标类型；确定所述目标类型对应的目标初始声学模型，所述目标初始声学模型的模型结构和所述目标类型匹配；基于目标语料对所述目标初始声学模型进行训练，得到声学模型；基于所述声学模型，确定语音识别模型。本发明可以提高语音识别模型的适应性。