In order to solve the problem of multi-sensor data synchronous acquisition and data noise elimination or reduction in electric energy metering box, this paper studies and designs a multi-feature data acquisition system. In this method, the MCU STM32F7 and STM32F1 are used as the main-auxiliary processor, and the multi-sensor data filtering and extended IO resource acquisition system are designed. Secondly, a standardized communication protocol is designed and adopted to realize synchronous acquisition of sensor data. Finally, an adaptive filtering method based on LMS is adopted to deal with and analyze the noise reduction of the collected data. In this paper, the temperature sensor ADT7320 is taken as an example to conduct data acquisition and filtering processing experiments. After data noise reduction processing, the noise with the frequency above 1Hz is basically eliminated and the noise less than 1Hz is basically reduced to 50-90%. Comparing with the traditional low-pass filter, the method in this paper can perform filtering adaptively according to the noise frequency, so that it can obtain the relatively pure and accurate sensor data.