基于Tacotron模型的油田射孔语音合成方案

Perforation Speech Synthesis Scheme in Oilfield Based on Tacotron Model

射孔施工作业是油田开采的关键步骤之一。针对油田射孔操作依赖井场人工操控的局限,将语音合成技术应用到油田射孔操作中,实现了射孔设备远程操作。在油田射孔操作指令数据集的基础上,实现姿势指令与操作指令的映射,并使用Tacotron模型的深度学习算法设计语音合成系统,通过梅尔频谱、梅尔倒谱距离等的实验对比分析,结果表明方案的有效性。

Perforating construction is one of the key steps in oilfield exploitation. Aiming at the limitation of oilfield perforation operation relying on manual control of the wellsite,the speech synthesis technology is applied to the oilfield perforation operation,realizing the remote operation of perforating equipment.This paper studies the mapping of posture instructions and operation instructions based on the oilfield perforation operation instruction data set,and uses the Tacotron model’s deep learning algorithm to design a speech synthesis system.Through experiments on Mel spectrum and Mel cepstrum distance,etc.The comparison result shows the effectiveness of the scheme.