一种基于飞机音频的机型智能识别方法

• Format: Patent
• Language: Chinese
• Published: 20.06.2023
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

一种基于飞机音频的机型智能识别方法。所述识别方法流程包括：采集不同型号飞机的音频，将音频进行预处理，提取音频的梅尔倒谱系数特征，训练残差自注意力模型；将目标飞机音频信号输入到训练好的模型中，得到目标飞机的型号。本发明的残差自注意力模型训练网络包括：嵌入层、位置编码器、残差自注意力模块、池化层和全连接层；嵌入层将特征进行降维，位置编码器对特征进行位置编码，残差自注意力模块进行特征学习，池化层主要防止网络发生过拟合现象，全连接层表示飞机型号的类别。本发明为飞机型号识别提供了一种高效准确识别方法，可以有效地提高飞机型号识别的准确性和鲁棒性。 The invention discloses an airplane type intelligent identification method based on airplane audio. The identification method comprises the following steps of collecting audios of different types of airplanes, preprocessing the audios, extracting Mel-frequency cepstrum coefficient features of the audios, and training a residual self-attention model, and inputting the audio signal of the target airplane into the trained model to obtain the model of the target airplane. The residual self-attention model training network comprises an embedding layer, a position encoder, a residual self-attentionmodule, a pooling layer and a full connection layer, wherein the embedding layer carries out dimension reduction on features, the position encoder carries