Audio Sampling Rate Research Articles

Speech Audio Modeling by Means of Causal Moving Average Equipped Gated Attention

In the paper we compare different attention mechanisms on the task of audio generation using unsupervised approaches following previous work in language modeling. It is important problem, as far as speech synthesis technology could be used to convert textual information into acoustic waveform signals. These representations can be conveniently integrated into mobile devices and used in such applications as voice messengers or email apps. Sometimes it is difficult to understand and read important messages when being abroad. The lack of appropriate computer systems or some security problems may arise. With this technology, e-mail messages can be listened quickly and efficiently on smartphones, boosting productivity. Apart from that, it is used to assist visually impaired people, so that, for instance, the screen content can be automatically read aloud to a blind user. Nowadays, home appliances, like slow cookers can use this system too for reading culinary recipes, automobiles for voice navigation to the destination spot, or language learners for pronunciation teaching. Speech generation is the opposite problem of automatic speech recognition (ASR) and is researched since the second half of the eighteen's century. Also, this technology also helps vocally handicapped people find a way to communicate with others who do not understand sign language. However, there is a problem, related to the fact that the audio sampling rate is very high, thus leading to very long sequences which are computationally difficult to model. Second challenge is that speech signals with the same semantic meaning can be represented by a lot of signals with significant variability, which is caused by channel environment, pronunciation or speaker timbre characteristics. To overcome these problems, we train an autoencoder model to discretize continuous audio signal into a finite set of discriminative audio tokens which have a lower sampling rate. Subsequently, autoregressive models, which are not conditioned on text, are trained on this representation space to predict the next token, based on previous sequence elements. Hence, this modeling approach resembles causal language modeling. In our study, we show that unlike in the original MEGA work, traditional attention outperforms moving average equipped gated attention, which shows that EMA gated attention is not stable yet and requires careful hyper-parameter optimization.

Developmental History of the Vocal Tract Organ

The Vocal Tract Organ has had a number of iterations resulting from advances in available technology as well as requirements of perceptual experiments and performance paradigms. The objective of this paper is to relate the development history of the Vocal Tract Organ from the original vision to what it is today as a modern version of the Vox Humana pipe organ stop for application in voice production and perception research. Descriptive METHODS/DESIGN: The latest Vocal Tract Organ is a polyphonic eight-channel eight-stop one manual Vocal Tract Organ that enables tab stop selected three-D printed vocal tracts to be used to create sound. This version includes eight stops (four for female vowel oral tracts and four for male vowel oral tracts). The stops are implemented using conventionally engraved pipe organ stop tabs labeled "Vox Humana Female" or "Vox Humana Male" followed by the three-D printed vowel: "EE", "AH", "ER" or "UU." This is described alongside the development stages from which it emerged and covers all previous versions of the Vocal Tract Organ. At the heart of the latest instrument is a Bela BeagleBone Black with a Bela cape audio expander board which incorporates eight 16-bit audio outputs at 44.1 kHz sampling rate (earlier versions based on the Arduino Mega board were limited to 8-bit audio at an audio sampling rate of 16.384 kHz which limited the overall output spectrum). The latest Vocal Tract Organ is programmed using the audio graphical programming language Pure Data which is directly compatible with the Bela system. The Pure Data patch creates eight larynx outputs at the pitches set by the keys depressed on the keyboard and these are routed to Vocal Tract Organ loudspeakers with three-D printed vocal tracts attached. The Bela system has enabled real-time synthesis of eight-note polyphonic sounds to eight separate three-D printed vocal tracts, each being selectable via an organ tab stop switch. The instrument has been cased in a purpose-designed and built prototype laser-cut enclosure that incorporates the eight tab stops, a MIDI keyboard input, a pipe organ style swell (volume) pedal connection, four stereo (eight channels) audio amplifiers and terminal connections for the eight loudspeakers. The Vocal Tract Organ functions as a musical instrument for performance and as an instrument for vowel and pitch perception research. Implementing it with the Bela family of processors allows for low audio latency of 1 ms and rapid prototyping due to being able to program directly with the high-level graphical audio programming language, Pure data (Pd).

机器翻译辅助的中蒙、维汉语音翻译数据集子集

At present, there are few public datasets for speech translation, especially those between Chinese and other low-resource languages. The development of end-to-end speech translation is limited by resources. In light of the research idea of international speech translation datasets, in this paper, we used the public speech recognition datasets (AISHELL and THUYG-20) to convert them into speech translation datasets through machine translation. After data processing, they were reviewed and verified by experts, so as to obtain high-quality speech translation datasets. The dataset includes Chinese-Mongolian speech translation dataset and Uygur-Chinese speech translation dataset, and the audio sampling rate is 16 kHz. The Chinese-Mongolian speech translation subset contains 1,919 items with a size of 238 MB. The Uygur-Chinese speech translation subset contains 3,692 samples with a size of 652 MB. This dataset can be used for the research on end-to-end speech translation, and provide data support for exploring the speech translation between Chinese and minority languages. As the dataset has been reviewed and verified by experts, it can also be combined with speech recognition dataset to study machine translation.

A CRNN System for Sound Event Detection Based on Gastrointestinal Sound Dataset Collected by Wearable Auscultation Devices

In this article, we set up a novel audio dataset named Gastrointestinal (GI) Sound Set which includes 6 kinds of body sounds Bowel sound, Speech, Snore, Cough, Groan, and Rub. We do sound event detection (SED) based on it, and can accurately detect 6 types of sound events. First, the GI Sound Set is collected by wearable auscultation devices. To ensure generalization, patients from five different hospital departments are recruited for data collection, along with a group of healthy subjects. GI Sound Set refers to Google AudioSet in data format but varies in audio length and sampling rate. Second, we extract Mel-filter features from the recordings and investigate the performance of different activation functions and neural network architectures for detecting sound events. We use data augmentation, class balance to deal with the problem of quantitative imbalance between classes on the dataset. We apply multiple instances learning(MIL) to give out not only bag-level results but also frame-level results. In this work, GI Sound Set is the largest body sound dataset to date, and our approach shows state-of-the-art performance with an average score of F1=81.06% evaluated on the test set. Due to its simple network and conventional processing method, our CRNN system has high universality, which can be used in other audio datasets, such as respiratory sound and heart sound.

Locating virtual sound sources at arbitrary distances in real-time binaural reproduction

A real-time system for sound spatialization via headphones is presented. Conventional headphone spatialization techniques effectively place sources on the surface of a virtual sphere around the listener. In the new system, sources can be spatialized at different distances from a listener by interpolating head-related impulse responses (HRIRs) measured between 20 and 160 cm. These HRIRs are stored in different databases depending on the audio sampling rate. To ease the real-time constraints, users can choose the number of HRIR taps used in the convolution, and an alternative interpolation technique (simplex interpolation) was implemented instead of trilinear interpolation. Subjective tests showed that such simplifications yield satisfactory spatialization for some angles and distances.

Simulated sound in advanced acoustic model videos

The Advanced Acoustic Model (AAM) and other time-step aircraft noise simulation models developed by Wyle can generate video animations of the noise environment. The animations are valuable for understanding details of noise footprints and for community outreach. Using algorithms developed by NASA, audio simulation for jet aircraft noise has recently been added to the video capability. Input data for the simulations consist of AAM’s one-third octave band sound level time history output, plus flight path geometry and ground properties. Working at an audio sample rate of 44.1 kHz and a sample “hop” period of 0.0116 s, a random phase narrow band sample is shaped to match spectral amplitudes. Ground reflection and low frequency oscillation are added to the hops, which are merged into a WAV file. The WAV file is then mixed with an existing animation generated from the same AAM run. The process takes place in near-real time, based on a location that a user selects from a site map. The presentation includes demonstrations of the results for a simple level flyover and for the departure of a high performance jet aircraft from an airbase.

Equal Error Rate and Audio Digitization and Sampling Rate for Speaker Recognition System

Equal Error Rate and Audio Digitization and Sampling Rate for Speaker Recognition System

Parametric Packet-Layer Model for Evaluation Audio Quality in Multimedia Streaming Services

We propose a parametric packet-layer model for monitoring audio quality in multimedia streaming services such as Internet protocol television (IPTV). This model estimates audio quality of experience (QoE) on the basis of quality degradation due to coding and packet loss of an audio sequence. The input parameters of this model are audio bit rate, sampling rate, frame length, packet-loss frequency, and average burst length. Audio bit rate, packet-loss frequency, and average burst length are calculated from header information in received IP packets. For sampling rate, frame length, and audio codec type, the values or the names used in monitored services are input into this model directly. We performed a subjective listening test to examine the relationships between these input parameters and perceived audio quality. The codec used in this test was the Advanced Audio Codec-Low Complexity (AAC-LC), which is one of the international standards for audio coding. On the basis of the test results, we developed an audio quality evaluation model. The verification results indicate that audio quality estimated by the proposed model has a high correlation with perceived audio quality.

From ITU-T G.722.1 to ITU-T G.722.1 Annex C: A New Low-Complexity 14kHz Bandwidth Audio Coding Standard

This paper describes the low-complexity 14kHz bandwidth audio coding algorithm which has been recently standardized by ITU-T as Recommendation G.722.1 Annex C (“G.722.1C”). The algorithm is an extension to ITU-T Recommendation G.722.1 and a doubled form of the G.722.1 algorithm to permit 14 kHz audio bandwidth using a 32 kHz audio sample rate, at 24, 32, and 48 kbit/s. The G. 722.1C codec features very high audio quality, extremely low computational complexity, and low algorithmic delay compared to other state-of-the-art audio coding algorithms. This codec is suitable for use in video conferencing and teleconferencing, and Internet streaming applications as well as a general-purpose 14 kHz audio codec. Subjective test results from the Characterization phase of G 722.1C are also presented in the paper.

Synthesis of Hand Clapping Sounds

We present two physics-based analysis, synthesis, and control systems for synthesizing hand clapping sounds. They both rely on the separation of the sound synthesis and event generation, and both are capable of producing individual hand-claps, or mimicking the asynchronous/synchronized applause of a group of clappers. The synthesis models consist of resonator filters, whose coefficients are derived from experimental measurements. The difference between these systems is mainly in the statistical event generation. While the first system allows an efficient parametric synthesis of large audiences, as well as flocking and synchronization by simple rules, the second one provides parametric extensions for synthesis of various clapping styles and enhanced control strategies. The synthesis and the control models of both systems are implemented as software running in real time at the audio sample rate, and they are available for download at at http://ccrma-www.stanford.edu/software/stk and http://www.acoustics.hut.fi/go/clapd

Analogue dialogue [3G mobile phone power saving

Power consumption remains a challenge in the design of 3G mobile phones. Although power-saving design is generally associated with the digital part of a multimedia terminal, careful partitioning between analogue and digital implementation can yield big power savings, particularly the audio and user-interface subsystems. This article suggests various power-saving strategies, illustrated by examples in the audio output, audio sample rate conversion, codecs, and touchscreen interface subsystems.

Proposed SMPTE Engineering Guideline: Emphasis of AES/EBU Audio in Television Systems and Preferred Audio Sampling Rate

Proposed SMPTE Engineering Guideline: Emphasis of AES/EBU Audio in Television Systems and Preferred Audio Sampling Rate

Ten Criteria for Evaluating Synthesis Techniques

A synthesis technique (Figure ib) is a strategy for generating sound samples based on some control information called parameters. Parameters generally change at a rate that is significantly slower than the audio sampling rate. Whereas sound samples are usually produced at a constant rate, parameter-setting messages are often sporadic and irregular. For example, in the familiar Yamaha DX7 synthesizer (Chowning and Bristow 1986), the synthesis technique is frequency modulation (FM) synthesis, which produces a steady stream of audio samples, whereas MIDI pitch-bend is a parameter that changes only when the performer moves the instrument's pitch wheel or foot pedal. For a historical summary of synthesis techniques, see, for example, Smith (1991). A processing technique (Figure la) is similar to a synthesis technique except that it includes an additional input of one or more audio sample streams at the audio sampling rate. An example of a common processing technique is reverberation. Here, the parameters are controls, such as the initial delay before the first reflections, the balance between the reverberant and dry signals, and the decay rate of the echoes. A wide variety of synthesis and processing techniques have been invented. The question may arise as to which technique is best. As it turns out, there is no simple answer. The best technique depends on the priorities of the user and the problem to be solved.

Editing Digital Audio Signals in a Digital Audio/Video System

The editing of video signals is usually based on the visual content of the video signal. This means the associated audio signal editing must be done to match the required final visual product. Since the visual editing is restricted at least to integer fields and more generally to every other frame, these same time restrictions will apply to the audio editing. When the video and audio signals are digitally encoded, their sampling rates should be related to each other exactly. This article shows how video systems may use the 50 kHz audio sample rate, how this rate is locked to the video system, and methods of performing the actual digital audio edits to coincide acceptably with the associated edited digital video signal.