ConvMixer：基于课程学习与特征交互卷积的小规模、嘈杂远场关键词识别
ConvMixer: Feature Interactive Convolution with Curriculum Learning for Small Footprint and Noisy Far-field Keyword Spotting
AI总结：
提出仅10万参数的ConvMixer模型，用特征交互卷积替代注意力，结合课程学习多条件训练，在嘈杂远场关键词识别中达到98.2%准确率，媲美大型Transformer。
摘要：
在神经语音处理中构建高效架构对于关键词识别部署的成功至关重要。然而，轻量级模型要利用简洁的神经操作实现噪声鲁棒性极具挑战。在现实应用中，用户环境通常嘈杂，且可能包含混响。我们提出了一种新颖的特征交互卷积模型，仅用10万参数即可应对嘈杂远场条件下的这一挑战。我们提出用交互单元替代注意力模块，以更高效的计算促进信息流动。此外，采用基于课程的多条件训练以获得更好的噪声鲁棒性。我们的模型在Google Speech Command V2-12上达到了98.2%的top-1准确率，并在设计的噪声条件下与大型Transformer模型相比具有竞争力。
英文摘要：
Building efficient architecture in neural speech processing is paramount to success in keyword spotting deployment. However, it is very challenging for lightweight models to achieve noise robustness with concise neural operations. In a real-world application, the user environment is typically noisy and may also contain reverberations. We proposed a novel feature interactive convolutional model with merely 100K parameters to tackle this under the noisy far-field condition. The interactive unit is proposed in place of the attention module that promotes the flow of information with more efficient computations. Moreover, curriculum-based multi-condition training is adopted to attain better noise robustness. Our model achieves 98.2% top-1 accuracy on Google Speech Command V2-12 and is competitive against large transformer models under the designed noise condition.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
