摘要
针对传统的单片机硬件系统复杂、处理能力弱、时间延迟大等问题,文章提出了一种基于智能AI芯片K210的声音采集与识别系统设计方法。该方法首先通过音频传感模块采集音频信号,其次通过主控模块提取MFCC特征并将其作为其声音特征,涉及预处理、快速傅里叶变换、Mel滤波器组、对数运算、离散余弦变换、动态特征提取等步骤。MFCC在降低信噪比方面具有良好的识别性能,能较好地利用听觉模型的成果。通过测试,文章验证了该方法的可行性及高速性,有效解决了传统方法速率低、成本高等问题。
This article proposes a design method for a sound acquisition and recognition system based on the intelligent AI chip K210 to address the problems of complex hardware systems,weak processing capabilities,and large time delays in traditional single-chip microcontrollers.This method first collects audio signals through an audio sensing module,and then extracts MFCC features through the main control module as its sound features,involving preprocessing,fast Fourier transform,Mel filter bank,logarithmic operation,discrete cosine transform,dynamic feature extraction and other steps.MFCC has good recognition performance in reducing signal-to-noise ratio and can effectively utilize the results of auditory models.Through testing,the article has verified the feasibility and speed of this method,effectively solving the problems of low speed and high cost in traditional methods.
作者
孙洪涛
刘春防
李望荣
SUN Hongtao;LIU Chunfang;LI Wangrong(Goer Co.,Ltd.,Weifang,Shandong 261006,China;Weifang Goer Electronics Co.,Ltd.,Weifang,Shandong 261021,China)
出处
《计算机应用文摘》
2024年第8期99-101,共3页
Chinese Journal of Computer Application
关键词
声音采集
识别
K210
sound collection
identification
K210
