Speech Coding Based on Spectral Dynamics

Machine Learning for Multimodal Interaction

et al.

Self Cite

Abstract. This paper proposes an analysis technique for wide-band audio applications based on the predictability of the temporal evolution of Quadrature Mirror Filter (QMF) sub-band signals. The input audio signal is first decomposed into 64 sub-band signals using QMF decomposition. The temporal envelopes in critically sampled QMF sub-bands are approximated using frequency domain linear prediction applied over relatively long time segments (e.g. 1000 ms). Line Spectral Frequency parameters related to autoregressive models are computed and quantized in each frequency sub-band. The sub-band residuals are quantized in the frequency domain using a combination of split Vector Quantization (VQ) (for magnitudes) and uniform scalar quantization (for phases). In the decoder, the sub-band signal is reconstructed using the quantized residual and the corresponding quantized envelope. Finally, application of inverse QMF reconstructs the audio signal. Even with simple quantization techniques and without any sophisticated modules, the proposed audio coder provides encouraging results in objective quality tests. Also, the proposed coder is easily scalable across a wide range of bit-rates.

Section: Structure Of the Codecmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Frequency Domain Linear Prediction for QMF Sub-bands and Applications to Audio Coding

Machine Learning for Multimodal Interaction

et al.

Self Cite

“…A novel speech coding system, proposed recently [1], exploits the predictability of the temporal evolution of spectral envelopes of speech signal using Frequency Domain Linear Prediction (FDLP) [2,3]. Unlike [2], this technique applies FDLP to approximate relatively long segments of the Hilbert envelopes in individual frequency sub-bands.…”

Section: Introductionmentioning

confidence: 99%

Perceptually Motivated Sub-band Decomposition for FDLP Audio Coding

Text, Speech and Dialogue

et al.

Self Cite

Abstract. This paper describes employment of non-uniform QMF decomposition to increase the efficiency of a generic wide-band audio coding system based on Frequency Domain Linear Prediction (FDLP). The base line FDLP codec, operating at high bit-rates (∼136 kbps), exploits a uniform QMF decomposition into 64 sub-bands followed by sub-band processing based on FDLP. Here, we propose a non-uniform QMF decomposition into 32 frequency sub-bands obtained by merging 64 uniform QMF bands. The merging operation is performed in such a way that bandwidths of the resulting critically sampled sub-bands emulate the characteristics of the critical band filters in the human auditory system. Such frequency decomposition, when employed in the FDLP audio codec, results in a bit-rate reduction of 40% over the base line. We also describe the complete audio codec, which provides high-fidelity audio compression at ∼66 kbps. In subjective listening tests, the FDLP codec outperforms MPEG-1 Layer 3 (MP3) and achieves similar qualities as MPEG-4 HE-AAC codec.

Non-uniform Speech/Audio Coding Exploiting Predictability of Temporal Evolution of Spectral Envelopes

Text, Speech and Dialogue

et al.

Self Cite

Abstract. We describe novel speech/audio coding technique designed to operate at medium bit-rates. Unlike classical state-of-the-art coders that are based on short-term spectra, our approach uses relatively long temporal segments of audio signal in critical-band-sized sub-bands. We apply auto-regressive model to approximate Hilbert envelopes in frequency sub-bands. Residual signals (Hilbert carriers) are demodulated and thresholding functions are applied in spectral domain. The Hilbert envelopes and carriers are quantized and transmitted to the decoder. Our experiments focused on designing speech/audio coder to provide broadcast radio-like quality audio around 15 − 25kbps. Obtained objective quality measures, carried out on standard speech recordings, were compared to the state-of-the-art 3GPP-AMR speech coding system.