Developing Behavioral Markers of Social Function in Aging: A Methodology Using Pitch Mimicry

The Covid pandemic brought to the forefront the crucial role of social interactions for society at large and in gerontological practice. Social interactions play a paramount role in preserving cognitive reserve in older adults. They rely on neurobehavioral processes that are complex (engage large parts of the brain and demand integrity of multiple perceptuomotor, attentional, cognitive and memory functions). Pitch mimicry is a well-known and spontaneously arising social phenomenon that requires the integrity of numerous processes of the brain, and we hypothesize that it constitutes a potentially sensitive behavioral marker of neurodegeneration in Alzheimer’s Disease and Related Dementias (ADRD). We developed and validated a series of algorithms to parse verbal exchanges between people and quantify the level of mimicry that each exhibit with their partners. Those algorithms are based on silence thresholding, carefully parametrized CEPSTRAL algorithms for automatic pitch estimation and Synchrosqueezing Transform for validation. We introduce a theoretical model to compare our estimates of pitch mimicry with model’s expectations based on the null hypothesis that its neurobehavioral pathways retain their integrity. Our method will allow researchers to study the evolution of pitch mimicry in aging individuals and its sensitivity to diverse social contexts, including those preserving lasting social engagement. Our method will also allow us to test the hypothesis that Pitch Mimicry is a sensitive behavioral marker of dementia, a condition characterized by a breakdown in social relatedness.

Carrier Phase-Based Precise Heading and Pitch Estimation Using a Low-Cost GNSS Receiver

Attitude and heading estimation methods using the global navigation satellite system (GNSS) are generally based on multi-antenna deployment, where the installation space and system cost increase with the increase in the number of antennas. Since the single-antenna receiver is still the major choice of the mass market, we focus on precise and reliable heading and pitch estimation using a low-cost GNSS receiver. Carrier phase observations are precise but have an ambiguity problem. A single difference between consecutive epochs can eliminate ambiguity and reduce the measurement errors. In this work, a measurement model based on the time-differenced carrier phases (TDCPs) is utilized to estimate the precise delta position of the antenna between two consecutive epochs. Then, considering the motion constraint, the heading and pitch angles of a moving land vehicle can be determined by the components of the estimated receiver delta position. A threshold on the length of the delta position is selected to avoid large errors in static periods. To improve the reliability of the algorithm, the Doppler-aided cycle slip detection method is applied to exclude carrier phases with possible cycle slips. A real vehicular dynamic experiment using a low-cost, single-frequency GNSS receiver is conducted to evaluate the proposed algorithm. The experimental results show that the proposed algorithm is capable of providing precise vehicular heading and pitch estimates, with both the root mean square errors being better than 1.5°. This also indicates that the cycle slip exclusion is indispensable to avoid unexpected large errors.

Dual Attention Network for Pitch Estimation of Monophonic Music

The task of pitch estimation is an essential step in many audio signal processing applications. In this paper, we propose a data-driven pitch estimation network, the Dual Attention Network (DA-Net), which processes directly on the time-domain samples of monophonic music. DA-Net includes six Dual Attention Modules (DA-Modules), and each of them includes two kinds of attention: element-wise and channel-wise attention. DA-Net is to perform element attention and channel attention operations on convolution features, which reflects the idea of "symmetry". DA-Modules can model the semantic interdependencies between element-wise and channel-wise features. In the DA-Module, the element-wise attention mechanism is realized by a Convolutional Gated Linear Unit (ConvGLU), and the channel-wise attention mechanism is realized by a Squeeze-and-Excitation (SE) block. We explored three kinds of combination modes (serial mode, parallel mode, and tightly coupled mode) of the element-wise attention and channel-wise attention. Element-wise attention selectively emphasizes useful features by re-weighting the features at all positions. Channel-wise attention can learn to use global information to selectively emphasize the informative feature maps and suppress the less useful ones. Therefore, DA-Net adaptively integrates the local features with their global dependencies. The outputs of DA-Net are fed into a fully connected layer to generate a 360-dimensional vector corresponding to 360 pitches. We trained the proposed network on the iKala and MDB-stem-synth datasets, respectively. According to the experimental results, our proposed dual attention network with tightly coupled mode achieved the best performance.