Fernando Huenupán

Nestor Becerra Yoma, Jorge Wuth, Andres Pinto et al.

In this paper, an end-to-end based LSTM scheme is proposed to address the problem of volcano event localization without any a priori model relating phase picking with localization estimation. It is worth emphasizing that automatic phase picking in volcano signals is highly inaccurate because of the short distances between the event epicenters and the seismograph stations. LSTM was chosen due to its capability to capture the dynamics of time varying signals, and to remove or add information within the memory cell state and model long-term dependencies. A brief insight into LSTM is also discussed here. The results presented in this paper show that the LSTM based architecture provided a success rate, i.e., an error smaller than 1.0Km, equal to 48.5%, which in turn is dramatically superior to the one delivered by automatic phase picking. Moreover, the proposed end-to-end LSTM based method gave a success rate 18% higher than CNN.

José Novoa, Josué Fredes, Jorge Wuth et al.

This paper addresses the combination of complementary parallel speech recognition systems to reduce the error rate of speech recognition systems operating in real highly-reverberant environments. First, the testing environment consists of recordings of speech in a calibrated real room with reverberation times from 0.47 to 1.77 seconds and speaker-to-microphone distances of 0.16 to 2.56 meters. We combined systems both at the level of the DNN outputs and at the level of the final ASR outputs. Second, recognition experiments with the reverb challenge are also reported. The results presented here show that the combination of features can lead to WER improvements between 7% and 18% with speech recorded in real reverberant environments. Also, the combination at DNN-output level is much more effective than at the system-output level. However, cascading both schemes can still lead to smaller reductions in WER.

Juan Pablo Escudero, José Novoa, Rodrigo Mahu et al.

Reverberation and additive noise have detrimental effects on the performance of automatic speech recognition systems. In this paper we explore the ability of a DNN-based spectral feature mapping to remove the effects of reverberation and additive noise. Experiments with the CHiME-2 database show that this DNN can achieve an average reduction in WER of 4.5%, when compared to the baseline system, at SNRs equal to -6 dB, -3 dB, 0 dB and 3 dB, and just 0.8% at greater SNRs of 6 dB and 9 dB. These results suggest that this DNN is more effective in removing additive noise than reverberation. To improve the DNN performance, we combine it with the weighted prediction error (WPE) method that shows a complementary behavior. While this combination provided a reduction in WER of approximately 11% when compared with the baseline, the observed improvement is not as great as that obtained using WPE alone. However, modifications to the DNN training process were applied and an average reduction in WER equal to 18.3% was achieved when compared with the baseline system. Furthermore, the improved DNN combined with WPE achieves a reduction in WER of 7.9% when compared with WPE alone.