In VoIP applications, such as Interactive Voice Response and VoIP-phone conversation transcription, speech signals are degraded not only by environmental noise but also by transmission network quality, and distortions induced by encoding and decoding algorithms. Therefore, there is a need for automatic speech recognition (ASR) systems to handle integrated noise-network distorted speech. In this study, we present a comparative analysis of a speech-to-text system trained on clean speech against one trained on integrated noise-network distorted speech. Training an ASR model on noise-network distorted speech dataset improves its robustness. Although the performance of an ASR model trained on clean speech depends on noise type, this is not the case when noise is further distorted by network transmission. The model trained on noise-network distorted speech exhibited a 60% improvement rate in the word error rate (WER), word match rate (MER), and word information lost (WIL) over the model trained on clean speech. Furthermore, the ASR model trained with noise-network distorted speech could tolerate a jitter of less than 20% and a packet loss of less than 15%, without a decrease in performance. However, WER, MER, and WIL increased in proportion to the jitter and packet loss as they exceeded 20% and 15%, respectively. Additionally, the model trained on noise-network distorted speech exhibited higher robustness compared to that trained on clean speech. The ASR model trained on noise-network distorted speech can also tolerate signal-to-noise (SNR) values of 5 dB and above, without the loss of performance, independent of noise type.
In IP audio systems, audio quality is degraded by environmental noise, poor network quality, and encoding–decoding algorithms. Therefore, there is a need for a continuous automatic quality evaluation of the transmitted audio. Speech quality monitoring in VoIP systems enables autonomous system adaptation. Furthermore, there are diverse IP audio transmitters and receivers, from high-performance computers and mobile phones to low-memory and low-computing-capacity embedded systems. This paper proposes MiniatureVQNet, a single-ended speech quality evaluation method for VoIP audio applications based on a lightweight deep neural network (DNN) model. The proposed model can predict the audio quality independent of the source of degradation, whether noise or network, and is light enough to run in embedded systems. Two variations of the proposed MiniatureVQNet model were evaluated: a MiniatureVQNet model trained on a dataset that contains environmental noise only, referred to as MiniatureVQNet–Noise, and a second model trained on both noise and network distortions, referred to as MiniatureVQNet–Noise–Network. The proposed MiniatureVQNet model outperforms the traditional P.563 method in terms of accuracy on all tested network conditions and environmental noise parameters. The mean squared error (MSE) of the models compared to the PESQ score for ITU-T P.563, MiniatureVQNet-Noise, and MiniatureVQNet–Noise–Network was 2.19, 0.34, and 0.21, respectively. The performance of both the MiniatureVQNet–Noise–Network and MiniatureVQNet-Noise model depends on the noise type for an SNR greater than 0 dB and less than 10 dB. In addition, training on a noise–network-distorted speech dataset improves the model prediction accuracy in all VoIP environment distortions compared to training the model on a noise-only dataset.