A Scale for the Measurement of the Psychological Magnitude Pitch

By S. S. Stevens, J. Volkmann, E. B. Newman. 1937

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Xu Tan
2. Tri Dao
3. Wei-Ning Hsu
4. Yu Zhang
5. Dong Yu
6. Albert Gu
7. Bo Li
8. Ron J. Weiss

Citations in the corpus

• Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron (2018, methodology): “In this work, we decided to use the same perceptually-relevant summarization of the spectrum that (Wang et al., 2017a) does: the mel-warped spectrum (Stevens et al., 1937).”
• Adversarial Audio Synthesis (2018, methodology): “This representation is projected to 128 frequency bins equally spaced on the Mel scale (Stevens et al., 1937) from 40 Hz to 7800 Hz .”
• ABX-Discriminability Measures and Applications. (Mesures de discriminabilité ABX et applications) (2016, background): “bandwidth of the filters estimated from masking experiments (Bark scale [174]) and based on the perceptual distance between sounds with different pitches (Mel scale [175]).”
• A perceptual interference account of acquisition difficulties for non-native phonemes. (2003, methodology): “The formant frequencies were spaced equally on the mel scale (Stevens et al., 1937).”
• A Deep Learning Approach to Data-driven Parameterizations for Statistical Parametric Speech Synthesis (2014, background): “Human hearing is known to be logarithmic both in amplitude[21], and frequency[22].”
• BirdNET: A deep learning solution for avian diversity monitoring (2021, methodology): “We performed frequency compression using a mel scale with 64 bands and a break frequency at 1750 Hz — considerably above the original proposal (Stevens et al., 1937) to achieve approximate linear scaling up to 500 Hz.”
• Origins and Development of Generalized Magnitude Representation (2011, background): “Pitch is another dimension lacking intrinsic polarity [84].”
• An Overview of Deep-Learning-Based Audio-Visual Speech Enhancement and Separation (2020, background): “• The frequency resolution of the human auditory system is not uniform, showing a higher discrimination for low frequencies [235].”
• AI4COVID-19: AI enabled preliminary diagnosis for COVID-19 from cough samples via an app (2020, methodology): “We used the Mel spectrogram over a typical frequency spectrogram because the Mel scale in the Mel spectrogram has unequal spacing in the frequency bands and provides a higher resolution (more informative) in lower frequencies and vice versa, as compared to equally spaced frequency bands in normal sp”
• Synthesizing Audio with Generative Adversarial Networks (2018, unknown): “This representation is projected to 128 frequency bins equally spaced on the Mel scale (Stevens et al., 1937) from 40Hz to 7800Hz.”
• Does Audio Deepfake Detection Generalize? (2022, unknown): “We evaluate these architectures on constant-Q transform (cqtspec [29]), log spectrogram (logspec) and mel-scaled spectrogram (melspec [30]) features (all of them 513-dimensional).”
• Filterbank learning for deep neural network based polyphonic sound event detection (2016, background): “On the other hand, the empirical nature of mel scale leads to several handcrafted mel formulas [12], [13], [14], [15], [16], all of which provide only an approximation to the real human perception.”
• Look, Listen, and Learn More: Design Choices for Deep Audio Embeddings (2019, background): “Mel spectrograms are designed to capture perceptually relevant information more efficiently with less frequency bands compared to a linear spectrogram [29].”
• Text-to-Speech Synthesis (2009, background): “From this, a new scale was defined in which onemel equalled one thousandth of the pitch of a 1KHz tone [415].”
• Wavelet Networks: Scale Equivariant Learning From Raw Waveforms (2020, background): “[63] Stanley Smith Stevens, John Volkmann, and Edwin B Newman.”