­Background Recurrent neural networks (RNNs) that enable accurate automatic speech recognition (ASR) are large in size and have long short-term memory (LSTM) capabilities. Due to the large size of these networks, most speech recognition tasks are performed in the cloud servers, which requires constant internet connection, introduces privacy concerns,...

Removing interferences and improving quality of degraded speech via speech enhancement (SE) is important in many applications such as telecommunications, speech recognition, voice over IP, hearing aids, and more. One major class of SE techniques includes machine learning, which is effective at enhancing speech quality, however, they require complex...

Background In automatic speech processing systems, speaker diarization is a crucial front-end component for separating speech segments by speaker without a priori knowledge about speaker identities. The first phase of many state-of-the-art diarization techniques involves the conversion of original speech data into representative i-vectors, which are...

Most modern speech compression (coding) algorithms, for example those used in cell phones, deliver narrowband (=3200Hz) telephone (near toll) quality speech. This limits the naturalness and intelligibility of the speech signal. Although wideband coding technology is available for most networks, the cost of renewing the entire infrastructure to support...

In recent years, the proliferation of internet streaming applications has brought about the need for low-bit rate speech and audio coding methods. Several parametric designs such as the sinusoids+transients+noise model have been somewhat successful for speech and audio synthesis. Low-bit rate and streaming applications are restricted by only...