Este trabalho estuda o modelo de análise em componentes independentes (ICA) para misturas instantâneas, aplicado na separação de sinais de áudio. Três algoritmos de separação de misturas instantâneas são avaliados: FastICA, PP (Projection Pursuit) e PearsonICA; possuindo dois princípios básicos em comum: as fontes devem ser independentes estatisticamente e não-Gaussianas. Para analisar a capacidade de separação dos algoritmos foram realizados dois grupos de experimentos. No primeiro grupo foram geradas misturas instantâneas, sinteticamente, a partir de sinais de áudio pré-definidos. Além disso, foram geradas misturas instantâneas a partir de sinais com características específicas, também geradas sinteticamente, para avaliar o comportamento dos algoritmos em situações específicas. Para o segundo grupo foram geradas misturas convolutivas no laboratório de acústica do LPS. Foi proposto o algoritmo PP, baseado no método de Busca de Projeções comumente usado em sistemas de exploração e classificação, para separação de múltiplas fontes como alternativa ao modelo ICA. Embora o método PP proposto possa ser utilizado para separação de fontes, ele não pode ser considerado um método ICA e não é garantida a extração das fontes. Finalmente, os experimentos validam os algoritmos estudados. / This work studies Independent Component Analysis (ICA) for instantaneous mixtures, applied to audio signal (source) separation. Three instantaneous mixture separation algorithms are considered: FastICA, PP (Projection Pursuit) and PearsonICA, presenting two common basic principles: sources must be statistically independent and non-Gaussian. In order to analyze each algorithm separation capability, two groups of experiments were carried out. In the first group, instantaneous mixtures were generated synthetically from predefined audio signals. Moreover, instantaneous mixtures were generated from specific signal generated with special features, synthetically, enabling the behavior analysis of the algorithms. In the second group, convolutive mixtures were probed in the acoustics laboratory of LPS at EPUSP. The PP algorithm is proposed, based on the Projection Pursuit technique usually applied in exploratory and clustering environments, for separation of multiple sources as an alternative to conventional ICA. Although the PP algorithm proposed could be applied to separate sources, it couldnt be considered an ICA method, and source extraction is not guaranteed. Finally, experiments validate the studied algorithms.

Made available in DSpace on 2016-08-17T14:52:55Z (GMT). No. of bitstreams: 1 Natalia Costa Leite Abreu.pdf: 841490 bytes, checksum: 00ff55b62f0819b502a66a2304564bf4 (MD5) Previous issue date: 2003-12-11 / The constant search for the improvement and strengthening of the relationship between humans and machines turning it more natural is common place. Consequently, the recognition of speech will turn, easier and practical the handling of equipments supplied with the capacity to understand the human speech. In this sense and with the use of the available knowledge information in the literature as how the human brain processes informations, some suggested methods try to simulate this ability in the computer, especially devoted to the extraction of a speech signal of mixed sounds, attempting, for example to increase the recognition and comprehension rate. The extraction of speech can be obtained by measures of a single-channel or multiple the channels. In order to extract the speech in a single channel, it is proposed here to use the speech characteristics introducing the concept of efficient codification, that tries to imitate the way the auditory cortex gets information using the method of Independent Component Analysis (ICA), getting the basis functions of the input signals and retrieving the estimated signal even when we add interferences to it. Our simulations also prove the efficiency of our method against reverberation effects and the recovery of speech signal by the handling of basis function of other speech signals. This technique can be used efficiently both to extract a single speech, as well as highlighting new ways of approaching the speech/speaker recognition problem. / A constante busca para aperfeiçoar e estreitar o relacionamento entre homens e máquinas, tornando-o mais natural, não é nenhuma novidade. Conseqüentemente, o reconhecimento da voz possibilitará uma manipulação mais fácil e prática de equipamentos dotados com a capacidade de compreender a fala humana. Neste sentido e utilizando-se dos conhecimentos disponíveis na literatura de como o cérebro humano processa informações, alguns métodos propostos procuram simular computacionalmente essa habilidade, voltados principalmente à extração de um sinal de voz de uma mistura de sons, na tentativa de, por exemplo, aumentar a taxa de reconhecimento e inteligibilidade. A extração da voz pode ser obtida usando medidas de um único ou múltiplos canais. Para extrair uma voz em um único canal, propomos usar as características da voz pelo conceito de codificação eficiente, que procura imitar o modo como o córtex auditivo trata as informações, utilizando-se da técnica de Análise de Componentes Independentes (ICA), obtendo as funções bases dos sinais de entrada e recuperando o sinal estimado, mesmo quando são adicionadas interferências. Através de simulações comprovamos também a eficiência da técnica usada, primeiro, na recuperação de um sinal de voz com a utilização das funções bases de outro sinal e, segundo, frente a efeitos de reverberação. Esta técnica pode ser usada para extrair uma única fala eficazmente, como também prenuncia um modo novo de chegar ao problema de reconhecimento da fala/orador.

The Short-Time Target Cancellation (STTC) algorithm, developed as part of this dissertation research, is a “Cocktail Party Problem” processor that can boost speech intelligibility for a target talker from a specified “look” direction, while suppressing the intelligibility of competing talkers. The algorithm holds promise for both automatic speech recognition and assistive listening device applications. The STTC algorithm operates on a frame-by-frame basis, leverages the computational efficiency of the Fast Fourier Transform (FFT), and is designed to run in real time. Notably, performance in objective measures of speech intelligibility and sound source segregation is comparable to that of the Ideal Binary Mask (IBM) and Ideal Ratio Mask (IRM). Because the STTC algorithm computes a time-frequency mask that can be applied independently to both the left and right signals, binaural cues for spatial hearing, including Interaural Time Differences (ITDs), Interaural Level Differences (ILDs) and spectral cues, can be preserved in potential hearing aid applications. A minimalist design for a proposed STTC Assistive Listening Device (ALD), consisting of six microphones embedded in the frame of a pair of eyeglasses, is presented and evaluated using virtual room acoustics and both objective and behavioral measures. The results suggest that the proposed STTC ALD can provide a significant speech intelligibility benefit in complex auditory scenes comprised of multiple spatially separated talkers. / 2020-10-22T00:00:00Z

At a cocktail party, one can choose to scan the room for conversations of interest, attend to a specific conversation partner, switch between conversation partners, or not attend to anything at all. The ability of the normal-functioning auditory system to flexibly listen in complex acoustic scenes plays a central role in solving the cocktail party problem (CPP). In contrast, certain demographics (e.g., individuals with hearing impairment or older adults) are unable to solve the CPP, leading to psychological ailments and reduced quality of life. Since the normal auditory system still outperforms machines in solving the CPP, an effective solution may be found by mimicking the normal-functioning auditory system. Spatial hearing likely plays an important role in CPP-processing in the auditory system. This thesis details the development of a biologically based approach to the CPP by modeling specific neural mechanisms underlying spatial tuning in the auditory cortex. First, we modeled bottom-up, stimulus-driven mechanisms using a multi-layer network model of the auditory system. To convert spike trains from the model output into audible waveforms, we designed a novel reconstruction method based on the estimation of time-frequency masks. We showed that our reconstruction method produced sounds with significantly higher intelligibility and quality than previous reconstruction methods. We also evaluated the algorithm's performance using a psychoacoustic study, and found that it provided the same amount of benefit to normal-hearing listeners as a current state-of-the-art acoustic beamforming algorithm. Finally, we modeled top-down, attention driven mechanisms that allowed the network to flexibly operate in different regimes, e.g., monitor the acoustic scene, attend to a specific target, and switch between attended targets. The model explains previous experimental observations, and proposes candidate neural mechanisms underlying flexible listening in cocktail-party scenarios. The strategies proposed here would benefit hearing-assistive devices for CPP processing (e.g., hearing aids), where users would benefit from switching between various modes of listening in different social situations. / 2022-05-19T00:00:00Z

It has been previously established that when engaged in a difficult attention intensive task, which involves repeating information while blocking out other information (the dichotic listening task), participants are often able to report hearing their own names in an unattended audio channel (Moray, 1959). This phenomenon, called the cocktail party effect is a result of words that are important to oneself having a lower threshold, resulting in less attention being necessary to process them (Treisman, 1960). The current studies examined the ability of a person who was engaged in an attention demanding task to hear and recall low-threshold words from a fictional story. These low-threshold words included a traditional alert word, "fire" and fictional character names from a popular franchise-Harry Potter. Further, the role of stimulus degradation was examined by including synthetic and accented speech in the task to determine how it would impact attention and performance. In Study 1 participants repeated passages from a novel that was largely unfamiliar to them, The Secret Garden while blocking out a passage from a much more familiar source, Harry Potter and the Deathly Hallows. Each unattended Harry Potter passage was edited so that it would include 4 names from the series, and the word "fire" twice. The type of speech present in the attended and unattended ears (Natural or Synthetic) was varied to examine the impact that processing a degraded speech would have on performance. The speech that the participant shadowed did not impact unattended recall, however it did impact shadowing accuracy. The speech type that was present in the unattended ear did impact the ability to recall low-threshold, Harry Potter information. When the unattended speech type was synthetic, significantly less Harry Potter information was recalled. Interestingly, while Harry Potter information was recalled by participants with both high and low Harry Potter experience, the traditional low-threshold word, "fire" was not noticed by participants. In order to determine if synthetic speech impeded the ability to report low-threshold Harry Potter names due to being degraded or simply being different than natural speech, Study 2 was designed. In Study 2 the attended (shadowed) speech was held constant as American Natural speech, and the unattended ear was manipulated. An accent which was different than the native accent of the participants was included as a mild form of degradation. There were four experimental stimuli which contained one of the following in the unattended ear: American Natural, British Natural, American Synthetic and British Synthetic. Overall, more unattended information was reported when the unattended channel was Natural than Synthetic. This implies that synthetic speech does take more working memory processing power than even an accented natural speech. Further, it was found that experience with the Harry Potter franchise played a role in the ability to report unattended Harry Potter information. Those who had high levels of Harry Potter experience, particularly with audiobooks, were able to process and report Harry Potter information from the unattended stimulus when it was British Natural. While, those with low Harry Potter experience were not able to report unattended Harry Potter information from this slightly degraded stimulus. Therefore, it is believed that the previous audiobook experience of those in the high Harry Potter experience group acted as training and resulted in less working memory being necessary to encode the unattended Harry Potter information. A pilot study was designed in order to examine the impact of story familiarity in the attended and unattended channels of a dichotic listening task. In the pilot study, participants shadowed a Harry Potter passage (familiar) in one condition with a passage from The Secret Garden (unfamiliar) playing in the unattended ear. A second condition had participants shadowing The Secret Garden (unfamiliar) with a passage from Harry Potter (familiar) present in the unattended ear. There was no significant difference in the number of unattended names recalled. Those with low Harry Potter experience reported significantly less attended information when they shadowed Harry Potter than when they shadowed The Secret Garden. Further, there appeared to be a trend such that those with high Harry Potter experience were reporting more attended information when they shadowed Harry Potter than The Secret Garden. This implies that experience with a franchise and characters may make it easier to recall information about a passage, while lack of experience provides no assistance. Overall, the results of the studies indicate that we do treat fictional characters in a way similarly to ourselves. Names and information about fictional characters were able to break through into attention during a task that required a great deal of attention. The experience one had with the characters also served to assist the working memory in processing the information in degraded circumstances. These results have important implications for training, design of alerts, and the use of popular media in the classroom.

No description available.

<pre><p>Difficulty understanding speech in background noise is the most common hearing complaint. Elucidating the neurophysiological mechanisms underlying speech intelligibility in everyday environments with multiple sound sources and distortions is hence important for any technology that aims to improve real-world listening. Using a combination of behavioral, electroencephalography (EEG), and computational modeling experiments, this dissertation provides insight into how the brain analyzes such complex scenes, and what roles different acoustic cues play in facilitating this process and in conveying phonetic content. Experiment #1 showed that brain oscillations selectively track the temporal envelopes (i.e., modulations) of attended speech in a mixture of competing talkers, and that the strength and pattern of this attention effect differs between individuals. Experiment #2 showed that the fidelity of neural tracking of attended-speech envelopes is strongly shaped by the modulations in interfering sounds as well as the temporal fine structure (TFS) conveyed by the cochlea, and predicts speech intelligibility in diverse listening environments. Results from Experiments #1 and #2 support the theory that temporal coherence of sound elements across envelopes and/or TFS shapes scene analysis and speech intelligibility. Experiment #3 tested this theory further by measuring and computationally modeling consonant categorization behavior in a range of background noises and distortions. We found that a physiologically plausible model that incorporated temporal-coherence effects predicted consonant confusions better than conventional speech-intelligibility models, providing independent evidence that temporal coherence influences scene analysis. Finally, results from Experiment #3 also showed that TFS is used to extract speech content (voicing) for consonant categorization even when intact envelope cues are available. Together, the novel insights provided by our results can guide future models of speech intelligibility and scene analysis, clinical diagnostics, improved assistive listening devices, and other audio technologies.</p></pre>

Sound source separation is a popular and active research area, especially with modern machine learning techniques. In this thesis, the focus is on single-channel separation of two speakers into individual streams, and specifically considering the case where two speakers are also accompanied by background noise. There are different methods to separate speakers and in this thesis three different methods are evaluated: the Conv-TasNet, the DPTNet, and the FaSNetTAC.  The methods were used to train models to perform the sound source separation. These models were evaluated and validated through three experiments. Firstly, previous results for the chosen separation methods were reproduced. Secondly, appropriate models applicable for NFC's datasets and applications were created, to fulfill the aim of this thesis. Lastly, all models were evaluated on an independent dataset, similar to datasets from NFC. The results were evaluated using the metrics SI-SNRi and SDRi. This thesis provides recommended models and methods suitable for NFC applications, especially concluding that the Conv-TasNet and the DPTNet are reasonable choices.