quantwave_core/indicators/
hamming.rs

1use crate::indicators::metadata::{IndicatorMetadata, ParamDef};
2use crate::traits::Next;
3use crate::utils::RingBuffer as VecDeque;
4use std::f64::consts::PI;
5
6/// Hamming Windowed FIR Filter
7///
8/// Based on John Ehlers' "Windowing" (S&C 2021).
9/// A finite impulse response (FIR) filter using a Hamming-like window with a pedestal for smoothing.
10#[derive(Debug, Clone)]
11pub struct HammingFilter {
12    length: usize,
13    _pedestal: f64,
14    window: VecDeque<f64>,
15    coefficients: Vec<f64>,
16    coef_sum: f64,
17}
18
19impl HammingFilter {
20    pub fn new(length: usize, pedestal_deg: f64) -> Self {
21        let mut coefficients = Vec::with_capacity(length);
22        let mut coef_sum = 0.0;
23
24        // Follows Ehlers' formula: Sine(Pedestal + (180 - 2*Pedestal)*count / (Length - 1))
25        // count from 0 to Length - 1
26        for count in 0..length {
27            let deg = pedestal_deg
28                + (180.0 - 2.0 * pedestal_deg) * count as f64 / (length as f64 - 1.0).max(1.0);
29            let coef = (deg * PI / 180.0).sin();
30            coefficients.push(coef);
31            coef_sum += coef;
32        }
33
34        Self {
35            length,
36            _pedestal: pedestal_deg,
37            window: VecDeque::with_capacity(length),
38            coefficients,
39            coef_sum,
40        }
41    }
42}
43
44impl Default for HammingFilter {
45    fn default() -> Self {
46        Self::new(20, 10.0)
47    }
48}
49
50impl Next<f64> for HammingFilter {
51    type Output = f64;
52
53    fn next(&mut self, input: f64) -> Self::Output {
54        self.window.push_front(input);
55        if self.window.len() > self.length {
56            self.window.pop_back();
57        }
58
59        if self.window.len() < self.length {
60            return input;
61        }
62
63        let mut filt = 0.0;
64        for (i, &val) in self.window.iter().enumerate() {
65            filt += self.coefficients[i] * val;
66        }
67
68        if self.coef_sum.abs() > 1e-10 {
69            filt / self.coef_sum
70        } else {
71            input
72        }
73    }
74}
75
76pub const HAMMING_FILTER_METADATA: IndicatorMetadata = IndicatorMetadata {
77    name: "HammingFilter",
78    description: "Hamming windowed FIR filter with pedestal.",
79    usage: "Apply as a windowing function before DFT-based cycle detection to reduce sidelobe leakage and obtain cleaner dominant cycle estimates.",
80    keywords: &["filter", "ehlers", "dsp", "windowing", "spectral"],
81    ehlers_summary: "The Hamming window is a raised-cosine weighting function that reduces spectral leakage by tapering the edges of a data block. Ehlers uses it in DFT-based cycle measurement tools to prevent energy in one frequency bin from contaminating adjacent bins, improving cycle period resolution.",
82    params: &[
83        ParamDef {
84            name: "length",
85            default: "20",
86            description: "Filter length",
87        },
88        ParamDef {
89            name: "pedestal",
90            default: "10.0",
91            description: "Pedestal in degrees",
92        },
93    ],
94    formula_source: "https://github.com/lavs9/quantwave/blob/main/references/traderstipsreference/TRADERS’ TIPS - SEPTEMBER 2021.html",
95    formula_latex: r#"
96\[
97Deg(n) = Pedestal + (180 - 2 \times Pedestal) \times \frac{n}{L-1}
98\]
99\[
100Coef(n) = \sin\left(\frac{Deg(n) \times \pi}{180}\right)
101\]
102\[
103Filt = \frac{\sum_{n=0}^{L-1} Coef(n) \cdot Price_{t-n}}{\sum Coef(n)}
104\]
105"#,
106    gold_standard_file: "hamming_filter.json",
107    category: "Ehlers DSP",
108};
109
110#[cfg(test)]
111mod tests {
112    use super::*;
113    use crate::traits::Next;
114    use proptest::prelude::*;
115
116    #[test]
117    fn test_hamming_basic() {
118        let mut ham = HammingFilter::new(20, 10.0);
119        for _ in 0..50 {
120            let val = ham.next(100.0);
121            approx::assert_relative_eq!(val, 100.0, epsilon = 1e-10);
122        }
123    }
124
125    proptest! {
126        #[test]
127        fn test_hamming_parity(
128            inputs in prop::collection::vec(1.0..100.0, 50..100),
129        ) {
130            let length = 20;
131            let pedestal = 10.0;
132            let mut ham = HammingFilter::new(length, pedestal);
133            let streaming_results: Vec<f64> = inputs.iter().map(|&x| ham.next(x)).collect();
134
135            // Batch implementation
136            let mut batch_results = Vec::with_capacity(inputs.len());
137            let mut coeffs = Vec::new();
138            let mut c_sum = 0.0;
139            for count in 0..length {
140                let deg = pedestal + (180.0 - 2.0 * pedestal) * count as f64 / (length as f64 - 1.0).max(1.0);
141                let coef = (deg * PI / 180.0).sin();
142                coeffs.push(coef);
143                c_sum += coef;
144            }
145
146            for i in 0..inputs.len() {
147                if i < length - 1 {
148                    batch_results.push(inputs[i]);
149                    continue;
150                }
151                let mut f = 0.0;
152                for j in 0..length {
153                    f += coeffs[j] * inputs[i - j];
154                }
155                batch_results.push(f / c_sum);
156            }
157
158            for (s, b) in streaming_results.iter().zip(batch_results.iter()) {
159                approx::assert_relative_eq!(s, b, epsilon = 1e-10);
160            }
161        }
162    }
163}
quantwave_core/indicators/hamming.rs

quantwave_core/indicators/
hamming.rs
