Neural Speech Phase Prediction Based on Parallel Estimation Architecture and Anti-Wrapping Losses

Abstract

This paper presents a novel neural speech phase prediction model which predicts wrapped phase spectra directly from amplitude spectra. The proposed model is a cascade of a residual convolutional network and a parallel estimation architecture. The parallel estimation architecture is a core module for direct wrapped phase prediction. This architecture consists of two parallel linear convolutional layers and a phase calculation formula, imitating the process of calculating the phase spectra from the real and imaginary parts of complex spectra and strictly restricting the predicted phase values to the principal value interval. To avoid the error expansion issue caused by phase wrapping, we design anti-wrapping training losses defined between the predicted wrapped phase spectra and natural ones by activating the instantaneous phase error, group delay error and instantaneous angular frequency error using an anti-wrapping function. We mathematically demonstrate that the anti-wrapping function should possess three properties, namely parity, periodicity and monotonicity. We also achieve low-latency streamable phase prediction by combining causal convolutions and knowledge distillation training strategies. For both analysissynthesis and specific speech generation tasks, experimental results show that our proposed neural speech phase prediction model outperforms the iterative phase estimation algorithms and neural network-based phase prediction methods in terms of phase prediction precision, efficiency and robustness. Compared with HiFi-GAN-based waveform reconstruction method, our proposed model also shows outstanding efficiency advantages while ensuring the quality of synthesized speech. To the best of our knowledge, we are the first to directly predict speech phase spectra from amplitude spectra only via neural networks.

A Dual-Branch Speech Enhancement Model with Harmonic Repair

“…On the DNS Challenge dataset, as seen in Table 4, our model improved PESQ without significantly increasing the number of model parameters and maintained an advanced level in other objective metrics. On the Voicebank + DEMAND dataset, as shown in Table 5, except for slight inferiority to DEMUCS [30] in CSIG, our model similarly shows significant improvements in all metrics. Compared to other studies, the advantage of our proposed method lies in considering the crucial role of harmonic information under noise masking for speech restoration.…”

“…The resulting output is up sampled to obtain the predicted CIRM spectrum's real part output M r . To avoid issues of non-structural and wrapping phase jumps [30], the Imag branch employs a phase decoder with dilated DenseNet [31]. After the deconvolution block, parallel dual 2D convolution layers output pseudo-components, and a dual-parameter arctangent function activates these two components to obtain the predicted CIRM imaginary spectrum output M i , where instance normalization layers are connected to standardize the network's intermediate features.…”