微型电脑应用

为了快速统计分析电力调度数据，明确电力调度状态，针对电力调度人员声纹特征提取精度和速率较低的问题，提出一种基于卷积神经网络的声纹特征AI回溯提取方法。运用模拟数字转换器获得调度人员语音数据信号序列，计算信号不同帧的短时能量与短时过零率均值，检测声纹信号端点，在Mel频率下依照人耳听觉特性，基于线性预测系数，采取非线性转换，通过微分、高阶微分等步骤获得声纹特征参数；运用AI技术中的卷积神经网络训练调度人员语音数据，引入连续片段说话人识别模型确定目标说话人身份，创建调度人员声纹判断函数，实现声纹特征AI回溯提取目标。仿真结果表明，所提方法声纹特征提取精度高、速率快，为电力调度系统的科学化管理提供重要决策思路。

In order to quickly analyze the power dispatching data and clarify the power dispatching state, aiming at the low accuracy and speed of voiceprint feature extraction of power dispatching personnel, an AI backtracking extraction method of voiceprint feature based on convolution neural network is proposed. The analog-to-digital converter is used to obtain the dispatcher's voice data signal sequence, calculate the short-time energy and short-time zero crossing rate mean value of different frames of the signal, detect the voiceprint signal endpoint. At Mel frequency, according to the human ear auditory characteristics, based on the linear prediction coefficient, nonlinear conversion is adopted to obtain the voiceprint characteristic parameters through differentiation, high-order differentiation and other steps. The convolution neural network in AI technology is used to train the dispatcher's voice data, the continuous segment speaker recognition model is introduced to determine the target speaker's identity, and the dispatcher's voiceprint judgment function is created to realize the AI backtracking extraction of voiceprint features. The simulation results show that the proposed method has high accuracy and fast speed of voiceprint feature extraction, and provides an important decision-making idea for the scientific management of power dispatching system.