Voices Obscured in Complex Environmental Settings (VOiCES) Corpus

Abstract

This paper introduces the Voices Obscured In Complex Environmental Settings (VOICES) corpus, a freely available dataset under Creative Commons BY 4.0. This dataset will promote speech and signal processing research of speech recorded by far-field microphones in noisy room conditions. Publicly available speech corpora are mostly composed of isolated speech at close-range microphony. A typical approach to better represent realistic scenarios, is to convolve clean speech with noise and simulated room response for model training. Despite these efforts, model performance degrades when tested against uncurated speech in natural conditions. For this corpus, audio was recorded in furnished rooms with background noise played in conjunction with foreground speech selected from the Lib-riSpeech corpus. Multiple sessions were recorded in each room to accommodate for all foreground speech-background noise combinations. Audio was recorded using twelve microphones placed throughout the room, resulting in 120 hours of audio per microphone. This work is a multi-organizational effort led by SRI International and Lab41 with the intent to push forward state-of-the-art distant microphone approaches in signal processing and speech recognition.

Robust Speaker Recognition Using Unsupervised Adversarial Invariance

“…V19-eval and V19-dev: We use the VOiCES data corpus [22] to evaluate the performance of our system with respect to the baselines on a speaker verification task and perform probing tasks to examine the systems. It consists of recordings collected from 4 different rooms with microphones placed at various fixed locations, while a loudspeaker played clean speech samples from the Librispeech [23] dataset.…”

“…Vox: Our training data consists of a combination of the development and test splits of VoxCeleb2 [25] and the development split [22]. Distractor represents noise source and green circles represents microphones Refers to number of sessions of VoxCeleb1 [26] datasets.…”

“…Consistent with general practice [6], equal error rate (EER) was used as the metric for evaluation. Following [22] and [27], we used knowledge of the nuisance factors annotations available in V19-eval dataset to study the various factors affecting the performance of speaker verification. For these experiments we consider two distinct nuisance factors, noise conditions: none, babble and television and microphone location: far-mic, near-mic and obstructed-mic (microphone hidden in the ceiling).…”

Structural Sparsification for Far-Field Speaker Recognition with Intel® Gna

“…For each data augmentation, we randomly choose from 2000 room impulse responses generated from Pyroomacoustics [21], and add randomly selected background noise from MUSAN [22] and AudioSet [23]. For the test set, we used the VOiCES far-field dataset [4], which we believe captures the essence of challenging channel conditions. For all speech utterances, we use 40-dimension log-mel filterbanks, with 3-second sliding window mean subtraction.…”

“…A number of speaker recognition systems based on deep neural network (DNN) embeddings have been reported in the literature [1][2] [3]. More recently, SRI developed the VOiCES dataset [4] specifically for far-field speaker recognition, and showed their DNN embeddings significantly outperformed the i-vector systems [5].…”