Search

This paper presents a complexity analysis of a spatial audio coding scheme (known as MPEG surrounding) based on a commercially available programmable device. The analysis mainly focuses on the complexity of spatial audio cue analysis and synthesis, especially for the additional special tools used during the spatial cue synthesis stage. The analysis results can be a reference during building up a SAC codec on a DSP.

© 2020 IEEE. A growing need for on-device machine learning has led to an increased interest in light-weight neural networks that lower model complexity while retaining performance. While a variety of general-purpose techniques exist in this context, very few approaches exploit domain-specific properties to further improve upon the capacity-performance trade-off. In this paper, extending our prior work [1], we train a network to emulate the behaviour of an audio codec and use this network to construct a loss. By approximating the psychoacoustic model underlying the codec, our approach enables light-weight neural networks to focus on perceptually relevant properties without wasting their limited capacity on imperceptible signal components. We adapt our method to two audio source separation tasks, demonstrate an improvement in performance for small-scale networks via listening tests, characterize the behaviour of the loss network in detail, and quantify the relationship between performance gain and model capacity. Our work illustrates the potential for incorporating perceptual principles into objective functions for neural networks.

Haptic communications allows physical interaction over long distances and greatly complements conventional means of communications, such as audio and video. However, whilst standardized codecs for video and audio are well established, there is a lack of standardized codecs for haptics. This causes vendor lock-in and thereby greatly limits scalability, increases cost and prevents advanced usage scenarios with multi-sensors/actuators and multi-users. The aim of this paper is to introduce a new approach for understanding and encoding tactile signals, i.e. the sense of touch, among haptic interactions. Inspired by various audio codecs, we develop a similar methodology for tactile codecs. Notably, we demonstrate that tactile and audio signals are similar in both time and frequency domains, thereby allowing audio coding techniques to be adapted to tactile codecs with appropriate adjustments. We also present the differences between audio and tactile signals that should be considered in future designs. Moreover, in order to evaluate the performance of a tactile codec, we propose a potential direction of designing an objective quality metric which complements haptic mean opinion scores (h-MOS). This, we hope, will open the door for designing and assessing tactile codecs.

During the recent years, high-level synthesis (HLS) has gained traction as a viable alternative to traditional handwritten register transfer level code in describing digital systems. This has been attributed to the maturing of the HLS tools and improving quality of their results. However, most published applications are data path intensive as HLS offers good tools for loop optimization, such as pipelining and loop unrolling. HLS is seldom applied to control-oriented applications since clock is not explicitly present in HLS source code. In this paper, we show how a clock cycle accurate application can be described with HLS. We give as a proof of concept an implementation of an FPGA-based I2C bus controller for an audio codec using Catapult C, and present a generalized work flow. Compared with a corresponding handwritten VHDL implementation, the HLS version consumes 84% more area at the same performance but productivity is increased by 100% at the first design time and even more with further design iterations. acceptedVersion

Recent trends in speech and audio codec standardization include scalability and extending the signal bandwidth beyond wideband (WB) to superwideband (SWB). In this paper we introduce a SWB extension for the ITU-T G.718 WB codec. In the SWB extension the high frequency content is generated utilizing the quantized MDCT domain coefficients of the WB core, which enables low additional delay. The proposed implementation is scalable with 4 kbps layers. In the first layer two different coding modes are used depending on the input signal type. The proposed SWB extension is evaluated with listening tests and complexity analysis.

This paper describes the speech and audio codec that has been submitted to ITU-T by Huawei and ETRI as a candidate for the upcoming super-wideband and stereo extensions of Rec. G.729.1 and G.718. The core codec in the current implementation is G.729.1 and the encoded frequency range is increased from 7 kHz to 14 kHz. Therefore, the maximum bit rate is raised from 32 kbit/s to 64 kbit/s by adding five bitstream layers. A comprehensive overview of the codec is presented with a focus on the mono coding components. The results of the listening tests that have been conducted during the ITU-T qualification phase are summarized. The proposed codec passes all quality requirements for mono input signals.

To help develop ultra-low power wireless hearing aid products, we investigate the integration of subband audio coding with hearing aid applications. Both the audio coding and the hearing aid application use subband processing, but their requirements for the filterbanks are totally different. The hearing aid application typically uses an oversampled filterbank to reduce the aliasing in each subband, whereas the audio codec needs a critically sampled filterbank for maximal coding efficiency. A joint filterbank structure is proposed in this paper to satisfy these contradictive filterbank requirements. With this structure, the two filterbanks are combined into a single stereo filterbank operation, which can be efficiently implemented on a filterbank coprocessor. This structure substantially reduces the computation complexity, power consumption and memory usage.

Audio coding based on Frequency Domain Linear Prediction (FDLP) uses auto-regressive model to approximate Hilbert envelopes in frequency sub-bands for relatively long temporal segments. Although the basic technique achieves good quality of the reconstructed signal, there is a need for improving the coding efficiency. In this paper, we present a novel method for the application of temporal masking to reduce the bit-rate in a FDLP based codec. Temporal masking refers to the hearing phenomenon, where the exposure to a sound reduces response to following sounds for a certain period of time (up to $200$ ms). In the proposed version of the codec, a first order forward masking model of the human ear is implemented and informal listening experiments using additive white noise are performed to obtain the exact noise masking thresholds. Subsequently, this masking model is employed in encoding the sub-band FDLP carrier signal. Application of the temporal masking in the FDLP codec results in a bit-rate reduction of about $10$\% without degrading the quality. Performance evaluation is done with Perceptual Evaluation of Audio Quality (PEAQ) scores and with subjective listening tests.

This paper describes a low-power implementation of the Bluetooth subband codec (SBC) for high-fidelity wireless audio. The design uses a configurable weighted overlap-add (WOLA) filterbank coprocessor to implement the analysis and synthesis filterbanks. A new method to convert the two-times over-sampled, complex WOLA subband signals to equivalent critically sampled, real-valued SBC subband signals is presented. The WOLA coprocessor allows for an efficient parallel implementation of the filterbank and quantization portions of the SBC algorithm. Details of the overall system design are also presented, including measurements of power consumption and resource requirements. The final real-time, fixed-point implementation is compared to an off-line floating-point reference and found to produce no audible difference in decoded signal quality.