The AISHELL-DMASH dataset was recorded in real smart home scenarios in two different rooms and contains 30,000 hours of speech data.The recording equipment includes a close-range microphone and 7 sets of equipment located in 7 different locations in the room. One set of recording equipment includes an iPhone, an Android phone, an iPad, a microphone, and a circular microphone array with a radius of 5 cm. The dataset contains 511 speakers, each of whom was visited 3 times with an interval of 7-15 days. The AISHELL-DMASH dataset is transcribed by professional speech annotators with a word accuracy of 98%, which can be used for research such as voiceprint recognition, speech recognition, and wake-up word recognition.