Role of non-linear data processing on speech recognition task in the framework of reservoir computing
Flavio Abreu Araujo, Mathieu Riou, Jacob Torrejon, Sumito Tsunegi, Damien Querlioz, K. Yakushiji, Akio Fukushima, Hitoshi Kubota, Shinji Yuasa, Mark D. Stiles, Julie Grollier
The reservoir computing neural network architecture is widely used to test hardware systems for neuromorphic computing. One of the preferred tasks for bench-marking such devices is automatic speech recognition (ASR). However, this task requires acoustic transformations from time domain sound waveforms to frequency domain maps that can be seen as feature extraction techniques. Depending on the method, these may obscure the effect of the hardware. Here, we quantify and separate the contributions of the acoustic transformations and the hardware. The non-linearity in the acoustic transformation plays a critical role in feature extraction as seen in the speech recognition success rate. We compute the gain in word success rate provided by a reservoir computing device, which is an appropriate bench-mark for comparing different hardwares.