Multi-view frequency-attention alternative to CNN frontends for automatic speech recognition

Convolutional frontends are a typical choice for Transformer-based Automatic Speech Recognition (ASR) to preprocess the spectrogram, reduce its sequence length, and combine local information in time and frequency similarly. However, the width and height of an audio spectrogram denote different information, e.g., due to reverberation as well as the articulatory system, the time axis has a clear left-to-right dependency. On the contrary, vowels and consonants demonstrate very different patterns and occupy almost disjoint frequency ranges. Therefore, we hypothesize, global attention over frequencies is beneficial over local convolution. We obtain 2.4 % relative word error rate reduction (rWERR) on a production scale Conformer transducer replacing its Convolutional Neural Network (CNN) frontend by the proposed F-Attention module on Alexa traffic. To demonstrate generalizability, we validate this on public LibriSpeech data with an Long Short Term Memory (LSTM)-based Listen Attend and Spell (LAS) architecture obtaining 4.6 % rWERR and demonstrate robustness to (simulated) noisy conditions.